Matching alcoholism treatments to client heterogeneity: Project

Matching AlcoholismTreatments to Client Heterogeneity:
Project MATCH PosttreatmentDrinking Outcomes
PROJECT
MATCH
RESEARCH
GROUP*
ABSTRACT. Objective:To assess
thebenefitsof matchingalcoholdependentclientsto threedifferenttreatments
with referenceto a variety
of client attributes.Method:Two parallelbut independent
randomized
clinical trials were conducted,one with alcoholdependentclientsreceivingoutpatient
therapy(N = 952;72% male)andonewithclientsreceivingaftercaretherapyfollowinginpatientor day hospitaltreatment
(N= 774; 80% male). Clients were randomly assigned to one
of three 12-week,manual-guided,individuallydeliveredtreatments:
CognitiveBehavioralCopingSkills Therapy,MotivationalEnhancementTherapyor Twelve-StepFacilitationTherapy.Clientswerethen
monitoredovera 1-yearposttreatment
period.Individualdifferencesin
response
to treatment
weremodeled
asa latentgrowthprocess
andevaluatedfor 10primarymatching
variablesand16contrasts
specified
a priori. The primaryoutcomemeasures
were percentdaysabstinentand
drinks per drinking day during the 1-year posttreatment
period.
Results:Clientsattendedon averagetwo-thirdsof treatmentsessions
offered,indicatingthat substantial
amountsof treatmentweredelivered,
and researchfollow-upratesexceeded90% of living subjectsinter-
viewedatthe1-yearposttreatment
assessment.
Significant
andsustained
improvementsin drinking outcomeswere achieved from base-
FTERANEXTENSIVE
review
ofalcoholism
outcome
research,the Institute of Medicine (1990) concluded
that it may no longerbe appropriateto ask whetheralcoholism treatment works or which treatment works best. Re-
flecting current views in the field, the report instead
suggested
thatthemoreimportantquestionis, "Whichkinds
of individuals,with what kinds of alcohol problems,are
likely to respondto what kindsof treatments
by achieving
whichkindsof goalswhendeliveredby whichkindsof practitioners?"(Instituteof Medicine,1990,p. 143).The "matching hypothesis"underlying this question assumesthat
prescribingspecifictreatments
basedon individualcharacteristicsandneedswould improvetreatmentoutcomescomparedto simplyofferingthesametreatmentto all individuals
with a similardiagnosis(DonovanandMattson,1994). The
Received:May 10, 1996.Revision:July3, 1996.
*ProjectMATCH is a collaborativeclinicaltrial sponsored
by the NationalInstituteon AlcoholAbuseandAlcoholism.The ProjectMATCH ResearchGroup is composedof the steeringcommitteememberswho
developed
thisresearch
protocolandexecuted
all aspects
of thetrial.Names
of the steeringcommitteemembersandcollaborating
institutions
appearin
the appendix.
Requestsfor reprintsshouldbe sentto the ScientificCommunications
Branch, National Institute on Alcohol Abuse and Alcoholism, Willco Build-
ing, Suite409, 6000 ExecutiveBlvd., Bethesda,MD 20892-7003.
line to 1-yearposttreatment
by the clientsassignedto eachof these
well-definedandindividuallydeliveredpsychosocial
treatments.
There
was little differencein outcomesby type of treatment.Only one attribute,psychiatricseverity,demonstrated
a significantattributeby
treatmentinteraction:
In theoutpatient
study,clientslow in psychiatric
severityhad more abstinentdaysafter 12-stepfacilitationtreatment
thanaftercognitivebehavioraltherapy.Neithertreatmentwasclearly
superior
forclientswithhigherlevelsof psychiatric
severity.
Two other
attributesshowedtime-dependent
matchingeffects:motivationamong
outpatients
andmeaning-seeking
amongaftercareclients.Clientattributesof motivationalreadiness,networksupportfor drinking,alcohol
involvement,
gender,psychiatric
severityand sociopathy
were prognosticof drinkingoutcomesovertime. Conclusions:
The findingssuggestthat psychiatricseverityshouldbe considered
whenassigning
clientsto outpatienttherapies.
The lackof otherrobustmatchingeffects
suggests
that,asidefrom psychiatric
severity,providersneednot take
theseclientcharacteristics
into accountwhentriagingclientsto oneor
the other of thesethree individuallydeliveredtreatmentapproaches,
despitetheir differenttreatmentphilosophies.
(J. Stud.Alcohol58:
7-29, 1997)
potentialbenefitsof treatmentmatchingincludeenhancement of treatmenteffectiveness,increasesin cost effectiveness, better utilization of resources, and avoidance of
therapeutic
mismatches
thatcouldcontributeto lack of responseto treatmentor dropoutfrom treatment(Finneyand
Moos, 1986; Institute of Medicine, 1990; Lindstrom, 1992;
Mattson and Allen, 1991; Miller, 1989).
The "matchinghypothesis,"
whichstatesthatclientswho
areappropriately
matchedtotreatments
will showbetteroutcomes than those who are unmatched or mismatched, is not
novel to medicine,behavioralscience(Beutler, 1979; Dance
and Neufeld, 1988; Keisler, 1966) or alcoholism treatment
(Bowmanand Jellinek,1941). The resultsof more than 30
previous
alcoholism
treatment
matching
studies
andthepotentialfor moreeffectivelyandefficientlyappliedtreatment
interventions
havemadetreatmentmatchinganexcitingclinical researchinterest.Empirical researchto date (Longabaughet al., 1994;Mattsonet al., 1994),however,indicates
onlythatmatchingis a promising,
butnotyetfully realized,
strategy
forincreasing
alcoholism
treatment
effectiveness.
In
1989 the National Institute on Alcohol Abuse and Alco-
holism(NIAAA) initiateda national,multisite,randomized
clinical trial of alcoholismtreatmentmatchingentitled
MatchingAlcoholismTreatments
to Client Heterogeneity
(ProjectMATCH). The studywasdesigned
to address
many
of thelimitations
ofpriormatching
studies,
particularly
in the
8
JOURNAL
OF STUDIES
ON ALCOHOL
areaof statisticalpower,andto providea rigoroustestof the
mostpromisingmatchinghypotheses.
ProjectMATCH consisted
of two parallelbutindependent
treatmentmatchingstudies,onewith clientsrecruitedat five
outpatientsites,the otherat five siteswith clientswho receivedaftercaretreatmentfollowingan episodeof inpatient
or intensivedayhospitaltreatment.Use of two parallelstudiesprovideda basisfor simultaneous
replicationandallowed
an evaluationof the matchinghypothesesin two major settingswhereambulatorytreatmentis oftendelivered:outpatientclinicsandasa follow-upto residentialcare.The overall
objectiveof eachstudywasto determineif varioussubgroups
of alcoholdependentclientswould responddifferentlyto
three manual-guided,individually delivered treatments:
CognitiveBehavioralCopingSkillsTherapy(CBT) (Kadden
et al., 1992), Motivational EnhancementTherapy (MET)
(Miller et al., 1992), and Twelve-StepFacilitationTherapy
(TSF) (Nowinski et al., 1992). Since theoreticallyderived
matchinghypotheses
wereconsidered
morelikely to be supportedthanthosegeneratedthroughotherstrategies
(Finney
andMoos, 1986, 1989;Longabaugh,1986),clientmatching
variablesandthe specificprimaryhypotheses
werebasedon
theoreticalconsiderations
and prior empiricalfindings.An
extensivereview of matchingstudies(Mattsonet al., 1994)
wasusedto developthe a prioriprimarymatchinghypotheses.Ten clientcharacteristics
wereselectedasmatchingvariables: (1) severityof alcohol involvement(Edwardsand
Lader, 1994; Orford et al., 1976); (2) cognitiveimpairment
(Cooneyet al., 1991; Donovanet al., 1987; Kaddenet al.,
1989); (3) clientconceptuallevel (McLachlan, 1972, 1974);
(4) gender(CronkiteandMoos, 1984;Lyonset al., 1982);(5)
meaningseeking(Brown, 1993;Fowler, 1993;Glaser,1993;
PishkinandFrederick,1973;Propst,1980); (6) motivational
readinessto change(DiClementeand Hughes, 1990; DiClemente et al., 1991; Heather et al., 1993; Marlatt et al.,
1988); (7) psychiatricseverity(Kaddenet al., 1989; McLellan et al., 1983a,b);(8) socialsupportfor drinkingversusabstinence(Longabaughet al., 1993, 1995); (9) sociopathy
(Cooneyet al., 1991;Kaddenet al., 1989);and(10) typology
(Litt et al., 1992).•
Specificmeasuresfor assessing
eachof the ten primary
matchingvariableswerechosen,andanticipatedinteractions
with eachof the threeselectedtreatmentswere specifiedin
hypothesized
contrasts.Table 1 identifiesthe specificmeasure used-for each of the 10 client characteristics and sum-
marizesthe 16hypothesized
contrasts
involvingthe 10 client
characteristics.
For example,it waspredictedthatthehigher
the level of alcohol involvement, the better the outcomesfor
clientsin bothCBT andTSF comparedwith thosein MET,
sincebothCBT andTSF weremorecomprehensive
andintensivethanthe MET intervention.Clientswho had greater
psychiatricseveritywere expectedto havebetteroutcomes
in CBT comparedto thosein eitherTSF or MET, sinceCBT
taughtskills for copingwith socialand emotionalcuesto
drink.For the motivationhypothesis,
lowerlevelsof readi-
/ JANUARY
1997
nessto changewere predictedto be associated
with better
outcomes for clients in MET, a motivation enhancement intervention, when contrasted with clients in CBT, a skills-
basedintervention.Acrosshypotheses,
eachtreatmentwas
assumedto havematchingpotentialfor specificclientcharacteristics.Cognitivebehavioraltherapywas hypothesized
to be especiallyeffectivefor clientswith higheralcoholinvolvement,cognitiveimpairment,psychiatricseverity,sociopathyandsupportfor drinking,aswell asfor womenand
Type B2 alcoholics.Twelve-stepfacilitationwas hypothesizedto be especiallyeffectivefor clientswith greateralcohol involvement,cognitiveimpairment,meaningseeking,
sociopathyand supportfor drinking,and for Type B alcoholics.Motivationalenhancement
therapywashypothesized
to be moreeffectivefor clientswith high conceptuallevels
andlow readinessto change.
Hypothesis
teamsdetermined
thespecificcontrasts
to be
tested,whetherto includein the hypothesisonly two or all
threetreatments,andwhetherto hypothesize
morethanone
contrastper attribute) The selectedcontrastswere tested
withtwoprimarydependent
outcomemeasures:
percentdays
abstinent(PDA) andaveragenumberof drinksperdrinking
day(DDD). Thisreportpresents
theresultsof testsof theprimary matchinghypotheseson these drinking outcomes,
alongwith analysesof main effectsfor treatmenttype, client
matchingattributeand site differencesduringthe year followingcompletionof the 12-weektreatmentperiod.
Method
Subjects
Although Project MATCH consistedof two independent arms of investigation,referredto as the "outpatient"
and "aftercare"studies,every effort was made to keep
them as similar as possible.In the outpatientarm, subjects
were recruiteddirectly from the communityor from outpatienttreatmentcenters.In theaftercarearm,thetreatments
were offeredto subjectsfollowing completionof inpatient
or intensivedayhospitaltreatment.
The outpatientandaftercarearmsof the trial involvedidenticalrandomizationprocedures, assessmentinstruments,treatment procedures,
follow-up evaluations,matching hypothesesand analytic
techniques.
Subjectswere recruitedat nine clinical researchunits
(CRUs) thatwereaffiliatedwith multipletreatmentfacilities.
The five outpatientCRUs werelocatedin Albuquerque,
NM,
Buffalo, NY, Farmington,CT, Milwaukee,WI,4 and West
Haven, CT. The aftercareCRUs were locatedin Charleston,
SC, Houston,TX, Milwaukee,WI, Providence,RI, andSeattle,
WA. The sitesreflectgeographic
aswell asclientheterogeneity. Outpatientsitesrecruitedsubjects
fromoutpatient
clinics
anddirectlyfrom thecommunitythroughadvertisements.
Aftercaresitesincludedsubjects
whohadbeentreatedin private,
publicandDepartment
of VeteransAffairs(VA) facilities.
PROJECT MATCH
TABLE 1.
RESEARCH
GROUP
9
Summary
of hypothesized
contrasts
for eachprimarymatching
variable
Clientattribute
Alcohol involvement
Cognitiveimpairment
Conceptuallevel
Gender
Meaningseeking
Motivation
Psychiatricseverity
Sociopathy
Supportfor drinking
Typology
Measuredby
Hypothesized
effectsa
AlcoholUseInventory
(Wanberget al., 1977)
ShipleyInstituteof Living
Scale:Trails A andB (Shipley,
1940);Symbol-DigitModalities
(Smith, 1973)
Paragraph
Completion
Method
(Hunt et al., 1978)
Self-report
Purpose
in Life Scale(Cmmbaugh
and
Maholik, 1976);Seekingof
NoeticGoalstest(Crumbaugh,
1977)
Subsetof URICA (DiClemente
andHughes,1990)
AddictionSeverityIndex:
Psych.Severitycompositescore
(McLellan et al., 1980)
CaliforniaPsychological
Inventory-Socialization
Scale
(Gough,1975)
ImportantPeopleandActivities
Instrument(Clifford and
Longabaugh,1991)
Compositeindexd
[CBT,TSF]slope> MET slope
0
TSF slope> CBT slope
½
CBT slope> MET slope
TSF slope> MET slope
MET slope> TSF slope
Female (CBT mean-TSF mean) >
male (CBT mean-TSF mean)
TSF slope> [MET,CBT] slopeø
CBT slope> MET slope
CBT slope> MET slope
CBT slope> TSF slope
CBT slope> MET slope½
CBT slope> TSF slope
TSF slope> MET slope
CBT slope> MET slopec
TSF slope> MET slope
Type B ([CBT,TSF] meanMET mean) > Type A
([CBT,TSF] mean - MET mean)ø
aThehypothesized
contrasts
predictdifferences
in slopesof theregression
linesfor eachtreatment
onoutcomeasa
functionof clientattribute.With theexceptionof the genderandtypologyattributes
(whichtakeon onlydiscrete
values),all contraststake the form: The differencebetweenthe first treatmentandthe secondbecomesmorepositive(orlessnegative)
withincreasing
valuesontheattribute.
Thegenderandtypologyattributes
taketheform:The
differencein meansbetweenthetreatments
is greaterat onelevelof theattributethanat theother.Hypotheses
did
not test whether interactions were ordinal or disordinal.
•Therationaleunderlying
the alcoholinvolvement,
meaningseekingandtypologyhypotheses
assumes
that,pertinentto the putativeactiveingredients
involvedin thehypothesized
matchingeffect,two treatments
are notdifferentin theireffect.Therefore,they werecombinedinto a singleconditionwhichwasthencontrasted
with the
third treatment.
cCognitiveimpairmentand sociopathy
each involvedthreehypothesized
treatmentcontrasts.
Therefore,the
Bonferronifamilywidecorrection
wasappliedto dividethe alphalevelby 3 for eachof theseattributes.
Support
for drinkingandpsychiatric
severityeachinvolvedtwohypothesized
contrasts;
thuseachof thesecontrasts
involved
dividingthealphalevelby 2. All otherattributesinvolvedsinglecontrasts.
al'he composite
typologyindexis derivedfrom severalinstruments
and sources.
For the purposes
of Project
MATCH, a five-variableindexwascomposed
of: percentof firstdegreerelativespositivefor alcoholdependence
takenfromtheAddictionSeverityIndex(McLellanet al., 1992);MacAndrewscalefromtheMMPI (MacAndrew,
1965);the totalscorefromthe EthanolDependence
Syndromescale(Babor,1996);Physicaleffectsof drinking
scorefrom theDrInC (Miller et al., 1995);andASP symptoms
takenfromthe Computerized
DiagnosticInterview
Schedule(CDIS) (Robinset al., 1989). Subjectswho scoredabovethe established
medianson threeof thesefive
scaleswereclassedasType B alcoholics(highvulnerability,highseverity).
Inclusion/exclusion
criteria. Inclusion criteria for the out-
patientstudywere:currentDSM-III-R diagnosisof alcohol
abuseor dependence;
alcoholastheprincipaldrugof abuse;
activedrinkingduringthe 3 monthspriorto entranceintothe
study;minimumageof 18; andminimumsixthgradereading
level.Exclusioncriteriawere:a DSM-III-R diagnosis
of current dependence
on sedative/hypnotic
drugs,stimulants,cocaine or opiates;any intravenousdrug use in the prior 6
months;currentlya dangerto selfor others;probation/parole
requirements
thatmightinterferewith protocolparticipation;
lackof clearprospects
forresidential
stability;inabilitytoidentify atleastone"locator"person
toassist
in trackingforfollowup assessments;
acutepsychosis;
severeorganicimpairment;
orinvolvement
(currentorplanned)in alternative
treatment
for
alcohol-related
problemsotherthanthatprovidedby Project
MATCH (definedasmorethan6 hoursof nonstudytreatment,
exceptfor self-helpgroupssuchas AlcoholicsAnonymous
[AA], duringthe3 monthsof studytreatment).
Criteria for the aftercare arm were identical, with the fol-
lowing modifications:DSM-III-R symptomsof alcohol
abuseor dependence
and requisitedrinkingbehaviorwere
assessed
for the3 monthspriorto theinpatientor dayhospital admission;
completion
of a programof at least7 daysof
inpatientor intensiveday hospitaltreatment(not simply
detoxification);
andreferralfor aftercaretreatment
by theinpatientor dayhospitaltreatmentstaff.
10
JOURNAL
OF STUDIES
ON ALCOHOL
/ JANUARY
1997
womenin thatarmof thestudy.In general,subjects
recruited
intothetwo studyarmsdifferedin predictableways:theoutpatientsampletendedto be significantlyyounger,moreresidentially stableand less dependenton alcoholthan the
aftercaresample(Goodmanet al., 1992;Timko et al., 1993).
A smallerproportionof outpatients(45%) than aftercare
clients(62%) reportedprioralcoholismtreatment.The overwhelmingnumberof clientsin eacharm (95% in outpatient,
98% in aftercare)met the criteriafor alcoholdependence
as
opposedto alcoholabuse,as assessed
usingthe Structured
Clinical Interview for DSM-III-R (Spitzer and Williams,
1985). Althoughindividualsdependenton other drugs(exceptfor marijuana)wereexcludedfromthetrial,therewasa
sizableminorityof subjectswho reportedsometypesof illicit drugusein the 90 dayspriorto recruitment.
In theoutpatient arm about 44% (n = 417) of the clients reported
someuseof illicit drugs,with men(46%) reportinga higher
low. For marijuana,the mediandaysof useof marijuanawas
low (rangingfrom 1 day duringthe 90-daypretreatment
period for aftercarewomento 4 daysfor outpatientmen).
Samplerepresentativeness.
In orderto recruita heterogeneoussample,a broad-based
recruitmenteffort was undertaken in multiple sites at CRUs. An initial screening
interviewwas conductedwith 2,193 potentialparticipants
for theoutpatientstudyand2,288for theaftercarestudy.Not
includedin thesefiguresare clientswho couldbe identified
(for example,throughchart review) as clearly ineligible
(e.g., primarydependence
on drugsotherthanalcohol)and
not administered
the screeninginterview.During the initial
screening,459 potentialparticipants(49 in outpatientand
410 in aftercare)indicatedthat they were not interestedin
participating.The major reasonscited for not taking part
werelogistical:45% mentionedtheinconvenient
locationof
the studyor transportation
problems,21% indicatedthattoo
muchtime was required,17% reportedthat theyplannedto
relocateand 16% statedthattheypreferredsomeothertreatmentoptionnotofferedin ProjectMATCH. Of the remaining 2,144 potential outpatient participants and 1,878
potentialaftercareparticipants,
952 (44%) wererandomized
in theoutpatientarm and774 (41%) wererandomizedin the
aftercarearm.Primaryreasons
for ineligibilitywere:failure
to completetheassessment
battery;residential
instability;legal or probationproblemsthat preventedrandomizationto
rate of use than women (39%). In the aftercare arm about
treatment
orprotocol
compliance;
comorbid
diagnosis
pre-
32% (n = 247) of theclientsreportedpretreatment
useof an
illicit drug,with women(36%) reportinga higherratethan
men (31%). However, frequencyof useof other drugswas
empting alcoholismtreatment;anticipationof concurrent
therapyin excessof thatpermittedin ProjectMATCH; failure to meetDSM-III-R criteriafor alcoholabuseor depen-
Other general admissionrequirementsfor all subjects
were:willingnessto acceptrandomization
to anyof thetreatment conditions;residencewithin reasonablecommuting
distance,with availabletransportation
to sessions;
andcompletionof priordetoxificationwhenmedicallyindicated.
Subjectcharacteristics.Table 2 describesthe characteristics of the 952 outpatients(72% male) and 774 aftercare
(80% male) clientsrecruited.Three of the five aftercaresites
were VA medical centers, which restricted recruitment of
TABLE2. Clientpersonalanddemographic
informationfor outpatient
andaftercarestudies
Aftercare
Outpatient
Variable
Gender
Men
Women
Total
Men
Women
Total
688
264
952
619
155
774
(72%)
38.7 -+ 10.5
(28%)
39.3 _+11.2
38.9 - 10.7
(80%)
42.0 _+10.9
(20%)
41.7 -+ 12.1
41.9 -+ 11.1
White
81
78
80
80
83
80
Black
4
9
6
15
13
15
13
10
12
4
3
3
Age(mean_+SD)
Ethnicity(%)
Hispanic
Other
Years of formal education
2
13.4 -+ 2.2
3
13.6 _+ 2.1
2
13.4 _ 2.2
1
13.1 - 2.0
1
13.1 _+ 2.2
1
13.1 -+ 2.1
(mean _+SD)
Relationship
status(%)
Couple
Single
Employmentstatus(%)
Employed
Not employed
38
62
29
71
36
64
35
65
29
71
34
66
56
44
38
62
51
49
49
51
45
55
48
52
48
52
39
61
45
55
64
36
52
48
62
38
5.8 -+ 1.9
5.6 -+ 1.9
5.8 -+ 1.9
6.9 _+1.8
6.4 _ 2.0
6.8 -+ 1.9
.18 -+ .19
.22 - .19
.19 _+.19
.21 _ .20
.31 - .23
.23 - .21
Prior alcoholtreatment(%)
Yes
No
Alcoholdependence
symptoms
(mean ñ SD)a
ASI psychiatric
severity
b
(mean _ SD)
aMeasured
by theSCID for the90-dayperiodpriorto enrollment;
symptomcountsrangefrom 1 to 9.
t'Composite
scorederivedfromtheAddictionSeverityIndex;higherscores
indicatehigherlevelsof severity.
PROJECT
MATCH
RESEARCH
GROUP
11
Assessment
instruments
andprocedures
dencediagnosis;
andinabilityto providea "locator."A majority (67%) of the nonparticipants
had multiple reasons
citedfor exclusion.All randomized
participants
areincluded
Intakeassessments.
If an individualappearedto meetthe
in the analyses.
inclusioncriteriaduringthe initial screening,a diagnostic
Althoughit is difficult to ascertainthe representativeness evaluationinterviewwasscheduled
to exploreeligibilitycriof any sampleof alcoholicsseekingtreatment,thesedatainteria in greaterdetail.This sessionconsisted
of a brief dedicatethat (1) mostof the subjectswho passedthe initial
mographichistory;the alcohol,drag and psychoticscreen
sections of the Structured Clinical Interview for DSM-III-R
screenbut who werelaterexcludedfromparticipation
were
excludedappropriately
becausethey did not satisfythe in(SpitzerandWilliams, 1985);andthe legal,psychiatric
and
clusionor exclusioncriteria;and (2) amongthosefoundto
family history sectionsof the AddictionSeverityIndex
be eligiblefor participation,
refusalswereattributable
to lo(McLellanet al., 1992).Subjects
alsocompleted
a 60-minute
gisticalconsiderations
ratherthanpersonalfactors,suchas
batteryof self-administered
questionnaires.
motivation.It is unlikelythattheselogisticalproblemslimA subsequent
pretreatment
evaluationsessionfocusedon
ited our ability to draw inferencesaboutmatchingeffects,
drinkingbehaviorandprevioustreatmentexperiences.
Estinoris therereasonto believethattherecruitmentprocedures
matesof alcoholconsumption
wereobtainedby meansof the
Form 90 (Miller, 1996; Miller and Del Boca, 1994), an infailed to providea broadrangeof clientstypically seenin
thesetypesof clinicalsettings.
terview procedurecombiningcalendarmemorycuesfrom
time-line follow-back methodology(Sobell and Sobell,
Procedure
1992) and drinkingpattemestimationprocedures
from the
Comprehensive
Drinker Profile (Miller andMarlatt, 1984).
Subjectswererecruitedovera 2-yearperiodusinga variIn additionto estimating
alcoholconsumption
for eachof the
ety of strategies
aimedat maximizingsampleheterogeneity previous90 days,theForm90 elicitsinformationaboutdrug
(Zweben et al., 1994). Following an initial screeninginteruse, treatmentexperiences,incarcerationand involvement
view to evaluateinclusion/exclusion
criteria,subjectsprowith AA. Also administered
duringthissessionwereseveral
vided informedconsentand participatedin three intake
neuropsychological
measuresof cognitivefunctionand a
sessionscomprisedof personal interviews, computersecondpacketof self-reportquestionnaires.
assistedassessment
and completionof self-administered
Thefinalassessment
session,
thepsychological
evaluation,
questionnaires.
As a quality assurancemeasure,all interconsistedof socialsupportmeasuresand psychologicalasviewswereaudiotaped.
Bloodandurinesampleswere also
sessments,
includingtheComputerized
DiagnosticInterview
obtainedat intake(in hospitalsettings,
patientsgavepermisSchedule
(C-DIS) (Robinset al., 1989),for purposes
of idensionto accessthesedata)and,wherepossible,an interview
tifyinganxiety,moodandantisocial
personality
disorders.
wasconducted
with anindividualfamiliarwith thesubject's
On averagethe entireassessment
battery,includingselfdrinking(a collateral).For outpatientparticipants,
the basereportquestionnaires,
tookabout8 hoursto complete.A deline assessment included a medical evaluation to determine
miledlistingof themeasures
includedin thefull batteryhas
the needfor medicallysuperviseddetoxification.If sucha
beenprovidedby Connorset al. (1994). The measuresasneedwasindicated,clientsweredetoxifiedpriorto randomsociatedwith the primarymatchingvariablesare identified
in Table 1.
ization.Randomization
to treatmentwasperformedusinga
computerizedurn balancingprogramdesignedto minimize
Follow-upassessments.
Eachof thefive follow-upassessdifferenceson criticaldemographic
andmatchingvariables
mentsessions
includeda core setof procedures
and instruments. To facilitate data collection from collaterals and
amongsubjects
acrossthethreestudytreatments
in eacharm
(ProjectMATCH ResearchGroup,1993;Stoutet al., 1994).
follow-uptracking,availableinformationregardingthe resiIn fact, therewere no significantdifferenceson dependent
dencesandtelephonenumbersof the client,collateralinformeasuresor matchingvariablesby treatmentconditionat
mantsandpotential"locators"
wasreviewedandupdated.
The
baseline assessment.
follow-upversionof theForm90 wasadministered
usingthe
Followingrandomization,
treatmentlastedfor 12 weeks.
dateofthelastinterviewasa starting
point.TherewerealsoteleTherapysessions
werevideotaped
to assurequalitydelivery
phoneinterview(Form90-T) andquickfollow-upinterview
of treatmentandto providethedataneededfor a detailedin(Form 90-Q) versionsfor uncooperative
clients.If clients
vestigationof treatmentprocess(Carroll et al., 1994; Dimisseda follow-upsession,
theywereassisted
atthenextsesClemente et al., 1994). Follow-up assessmentswere
sionin reconstructing
theiralcoholconsumption
fortheprevischeduledat 3 (endof treatment),6, 9, 12 and 15 monthsafousperiod.Continuous
dailydrinkingestimates
wereproduced
terthefirsttherapysession.
The 3-, 9- and 15-monthsessions
for the entire 1-yearposttreatment
follow-upperiod.The
were major evaluationpoints,involving the collectionof
Drinker Inventoryof Consequences
(DrlnC) (Miller et al.,
1995)alsowasadministered
ateachof thefive follow-upevalbloodandurinespecimens,
andcollateralinterviews.
A more
completedescription
of thetrial protocolhasbeenprovided
uationsto assess
problems
associated
with alcoholuse.Other
by theProjectMATCH Research
Group(1993).
baselineassessment
instruments
wererepeated
at threemajor
12
JOURNAL
OF STUDIES ON ALCOHOL
assessment
points(3, 9 and15monthsfollowingentryintothe
study).
Collateralandbiochemicalmeasures.
Collateral informants
andlaboratory
testswereusedto monitorchanges
in subjects'
alcoholconsumption
andto corroborate
self-reportmeasures.
Blood sampleswere analyzedto monitor liver enzymes
(GGTP, SGOT, SGPT). Carbohydrate-deficient
transfertin
(CDT), a markerfor heavydrinking,wasassessed
in the 15monthbloodsample(AntonandBean,1994;AntonandMoak,
1994).Urinesamples
werescreened
forrecentuseof fivepsychoactivesubstances:
opiates,cannabinoids,
amphetamines,
benzodiazepines
andcocaine.CDT andurinespecimens
were
assayed
atacentrallaboratory
(ClinicalNeurobiology
Laboratory,MedicalUniversityof SouthCarolina,Charleston).
Completeness
and accuracyof data. For botharmsof the
study,datafor over90% of the subjectswerecollectedfor all
five (at 3, 6, 9, 12 and15months)follow-uppoints.Thisfigure includessubjectsfor whom data from an earlier time
pointwerereconstructed
at a laterfollow-up(thefrequency
of such reconstructions
for any given assessment
period
rangedfrom4-6% for outpatientparticipants
andfrom4-8%
for aftercareparticipants).
The Form90-T (telephone)interview wasusedinfrequentlyfor follow-updatacollection(the
ratesfor the follow-upsat 3, 6, 9 and 15 monthswere, respectively,3%, 8%, 6% and7% for theoutpatientstudyand
5%, 19%, 6% and 6% for the aftercarestudy).The Form
90-Q (quick)for uncooperative
clientswasalsousedrarely
(< 1%of theoutpatient
participants
and<2% of theaftercare
participants
at anygivenfollow-uppoint).At the 1-yearposttreatment(15-month)evaluationsession,93% of the living
aftercareclientsand92% of thelivingoutpatientclientswere
interviewed.Client deathsduring active treatment(n = 3)
and follow-up (n = 24) phasesof the trial totaled 1.6% of
thoserandomized.Blood sampleswere obtainedat 1-year
posttreatment
from 83% of the aftercareand82% of the outpatientclients.Urine sampleswereprovidedby 85% of the
clientsin eacharm of the study.Collateralinformantswere
/ JANUARY
1997
Toniganetal.,inpress).
Cross-site
reliabilities,
asindexed
by
theintraclass
correlation
of ratingsof thesameclientby researchassistants
fromdifferentsites,werehigh.
Urine drug screenswere highly consistentwith selfreporteddruguseat baselineandfollow-up.Whendiscrepancieswereobserved,
it wasmorelikelythatclientsreported
drugusewhentheurinescreenwasnegative.Similarlyclients
tendedto reportmoreuseof drugsandalcoholthandid their
collateralinformants.
Self-reports
of drinkingwerealsoexaminedin relationtogammaglutamyltranspeptidase
(GGTP)
valuesat the 15-monthassessment
point.Clientswerepartitionedintotwogroupsonthebasisof GGTPvaluesbeingnormal or abnormal:27% (30% of men, 20% of women) of
outpatientclientsand32% (33% of men, 25% of women)of
aftercareparticipantshad GGTP values in the abnormal
rangeat the 15-monthfollow-uppoint.GGTP Status(normal
vs abnormal)x GenderANOVAs were performedfor the
two primary drinkingmeasures(PDA and DDD) summed
overthe30-dayperiodpriorto blooddraw.Statistically
significantGGT maineffects(p < .05) were obtainedfor both
drinkingvariablesin botharmsof the trial, indicatingthat
self-reported
alcoholusewasconsistently
higherfor clients
with abnormalGGT testresults.
5 In a separate
analysis40%
of theaftercareclients(45% of menand25% of women)and
35% of theoutpatients
(42% of men,18%of women)hadabnormallyhigh CDT levelsindicativeof heavyalcoholconsumption.Clients who had CDT levels above the normal
cut-off(> 17u/1for menand> 25u/1for women)hadhigher
self-reported
drinkingasindicatedby lowerPDA andhigher
DDD (p values< .01 for bothdependent
measures
in both
armsof the trial) for the monthprior to the 15-monthinterview. Althoughgenderdifferences
needfurtherexploration,
therelationship
of CDT to theself-reportdataconfirms,in the
aggregate,the validity of the verbally reporteddrinking.
Takentogether,thereliabilityandvaliditydataindicatethat
a highdegreeof confidence
canbeplacedin theaccuracy
of
theverbalreportdataobtainedin ProjectMATCH.
contacted at baseline and at 3, 9 and 15 months and inter-
viewed usingthe collateralform of the Form 90. Contact
Treatments
for matching
rates for named collaterals at baseline were 87% and 83% in
the aftercareandoutpatientarms,respectively,anddeclined
to 78% and 75% at the 1-year posttreatmentevaluation.
Techniquesemployedto assuredata quality are described
elsewhere(ProjectMATCH ResearchGroup, 1993).
Reliabilityand validity of verbal reportmeasures.Given
theextentto whichverbalreportmeasures
werereliedupon
for subjecteligibilityand for assessment
of matchingvariablesandtreatmentoutcomes,
specialattentionwasgivento
the evaluationof reliability and validity of interviewsand
questionnaires
employedin the trial. A comprehensive
testretestreliabilitystudyshowedthatmeasures
derivedfrominterviewerassessments
werereliablefor interviewerspaired
bothacross
andwithinsites.In particular,
theForm90 primary
outcomemeasures
(PDA andDDD) werefoundto be consistent acrosstest-retestinterviews(Del Bocaand Brown, 1996;
Three treatmentmodalitieswere chosenas potential
matchesfor clientcharacteristics:
CognitiveBehavioralCoping SkillsTherapy(CBT), MotivationalEnhancement
Therapy(MET) andTwelve-StepFacilitation(TSF) (Donovanet
al., 1994). Treatmentswere selectedbasedon potentialfor
matching,evidenceof clinicaleffectiveness,
distinctiveness
fromotherProjectMATCH treatments,
feasibilityof implementation,and applicabilitywithin existingtreatmentsystems. Although alcoholismtreatmentis often deliveredin
groupformat,designandmethodological
considerations
led
the researchgroupto chooseindividuallydeliveredtreatments. All three treatments were delivered over a 12-week
period:CBT andTSF bothinvolvedweeklytreatmentsessions,whereasMET consistedof four sessions,
occurring
duringthefirst, second,sixthandtwelfthweeks.
PROJECT
MATCH
RESEARCH
Treatmentsdifferedfrom oneanotherin a numberof ways
(Donovanet al., 1994).CBT wasbasedon sociallearning
theoryandvieweddrinkingbehaviorasfunctionallyrelated
to major problemsin an individual'slife, with emphasis
placedon overcomingskillsdeficitsandincreasingthe ability to copewith situations
thatcommonlyprecipitaterelapse.
TSF wasgrounded
in theconcept
of alcoholism
asa spiritual
and medicaldiseasewith statedobjectivesof fosteringacceptance
of thedisease
of alcoholism,
developing
a commitmentto participate
in AA andbeginning
to workthroughthe
12 steps.It shouldbe notedthatthe TSF interventiondoes
not representa testof AA asa treatmentintervention,but is
insteada treatmentdesigned
to promotethe client'sbeginningto workonthe12stepsandfosteractiveparticipation
in
traditionalfellowshipactivitiesof AA. MET wasbasedon
principlesof motivationalpsychologyandfocusedon producinginternallymotivatedchange.This treatmentwasnot
designedto guidethe client,stepby step,throughrecovery,
butinsteademployedmotivationalstrategies
to mobilizethe
individual'sown resources.
The therapyprotocolfor each
modalityis describedin detailedtherapymanuals(Kadden
GROUP
13
outsidetheProjectMATCH protocol.Thisconstituted
3.3%
(n = 57) of therandomized
sample,with no significant
differences in the numbers affected across treatment conditions.
Retentionand treatmentcompliance.
Clientsassigned
to
thethreeconditions
completed68% of theirscheduled
treatmentsessions
in theoutpatientand66% in theaftercaresites.
Directcomparisons
betweentreatments
aredifficultbecause
the MET
intervention
consisted of fewer sessions over the
12-weekperiodandTSF clientswere encouraged
to attend
AA meetingsin additionto the 12 individualtreatmentsessions.Analysescomparing
the threetreatments
in termsof
weeksin treatment(the numberof weeksthe client attended
treatment)revealedthatCBT clientsattendedtherapysignificantlylonger(9.3 weeks)thantheir MET (8.4) and TSF
(8.3) counterparts.
However,thiseffectwasobserved
onlyin
theoutpatientarmandtheeffectsizewassmall,a difference
of 1 week or less. In sum, clients received a substantial
amount of the tested treatments and differences in dose or
compliancebetweentreatmentswere small,suggesting
that
treatments
weredeliveredwith sufficientintensityandcomparabilityto testmatchinghypotheses.
et al., 1992; Miller et al., 1992; Nowinski et al., 1992).
A trainingprotocol
andstandards
fortherapist
certification
andmonitoring
weredeveloped.
Eightytherapists
werecerti-
Data analysisplan
fied to administer the three treatments in the trial. All sessions
Two primarydependentvariableswerechosenfor analysis. Percentdays abstinent(PDA) provideda measureof
drinkingfrequency.Drinksper drinkingday (DDD) constituteda measureof drinkingseverity(Baboret al., 1994).
Drinkingwas summarizedon a monthlybasis;if a person
wasabstinent
duringa givenmonth,hisor herscorefor the
werevideotaped
andsupervisors
monitored
25% of all Project
MATCH therapysessions
(over2,500)to ensuretherapist
adherencetotreatment
manualsandtopreventtherapistvariation
fromtheprotocol(ProjectMATCH Research
Group,1993).
Treatmentintegrity.Evaluations
of treatmentintegrityincludedtreatment
fidelityanddiscriminability,
treatment
dose,
exposureto nonstudytreatments
andlevel of therapistskillfulness(DiClementeet al., 1994). In botharmsof the study:
(1) studytreatments
wereimplemented
asintended,withhigh
discriminability
amongtreatments
basedonvideotape
ratings
of independentraters unawareof treatmentassignment;
(2) clientsreceivedsubstantial
exposureto studytreatments
withhighcontrastin treatmentexposure
betweensubjects
in
MET versusCBT andTSF; (3) exposure
to nonstudytreatments(excludingself-helpgroups)wasminimalandcomparableacrosstreatmenttypes;and(4) treatments
werelargely
comparable
with respectto nonspecific
dimensions
of the
treatment,suchastheworkingallianceandtherapistskillfulness(seeCarrollet al., submitted
for publication).
Clinical management
of subjects.Client progressduring
treatmentwas reviewedat therapistsupervisionmeetings
variable DDD was zero.
Individualdifferences
in response
toalcoholtreatment
were
modeledasa "latentgrowthprocess"
(BrykandRaudenbush,
1987).Therationalefor theselectionof thisapproach
for Project MATCH is providedelsewhere(Carbonaftet al., 1994).
The ?•oc MIXEDprocedure
of SASwasusedfor theseanalyses(SAS Institute,1992). Each subject'sgrowthcurveis a
polynomialfunctionof time.Quadraticlatentgrowthcurves
basedonpreliminarymodelfittinganalyses
wereused.Each
matchinghypothesis
wastestedseparately
at a family-wise
TypeI errorrateof 5%. If, for example,therewerethreehypotheses
relatingto a singlematching
variable,thenthosehypotheses
weretestedat a Bonferroni-corrected
alphalevelof
.05/3.Because
thereweretwo dependent
variables,thealpha
levelwasfurthercorrected
by a factorof 2. A Bonferroni
correctionwasemployedwithin,butnotacross,hypothesis
fam-
held weekly at eachCRU. A clinicalcarereview committee
ilies because a correction across families would lead to an
provideduniformguidanceacrossthe CRUs on decisions
concerningclinical "deterioration"and removalof clients
from thetreatmentprotocols.Deterioration
criteriaincluded
suicidalor homicidalrisk,onsetof significantcognitiveimpairment,deterioration
of physicalhealthandneedfor long-
excessivelyconservative
testandinflatedType II errorrate.
term hospitalizationor other intensivetreatment.Clients
who deterioratedor were at seriousrisk despiteProject
MATCH
treatments were referred for additional intervention
Alcohol
outcome variables such as PDA
and DDD
are
proneto substantial
departures
from normalitybecauseof
skewness
andfloor/ceiling
effects.Preliminary
analyses
indicatedthatanarcsintransformation
for PDA anda squareroot
transformation
for DDD improvedthe distribution
of these
variables.Subjectswere excludedfrom the latent growth
analyses
if morethan4 of 12months
oftheirdrinkingoutcome
14
JOURNAL
OF STUDIES
ON ALCOHOL
dataweremissing.Missingdrinkingdatacaused8.0% of aftercaresubjectsand 7.1% of outpatientsto be dropped.A
smallnumberof additionalsubjectswereexcludedfrom some
matchingtestsbecauseof missingdataonthematchingvariablebeinganalyzed.Of the 1,596subjectswhohadadequate
outcomedatato be includedin theanalyses,6.8% hadat least
onemissingmonthlyoutcomepoint.In aggregate,however,
therewereonly 100 missingmonthlyoutcomepointsoutof
8,544, or 1.2%.Ancillaryanalysesindicatedthattheprimary
analysisresultswerenot sensitiveto themissingdataexclusionrule.Thiswasnotsurprising
sincetherateof missingdata
was low.
Latentgrowthanalysisincorporatingan intentto treatapproachthatincludedall randomizedclientswasusedto test
eachmatchingvariablefor an attributeby treatmentinteraction (ATI) andtwo timeby attributeby treatmenteffects(linear and quadratic)for eachof the two dependentvariables
(PDA and DDD). Theseanalyseswere adjustedusinga set
of covariates to control for extraneous effects. 6 The covari-
ate adjustmentreportedhereincludedthe baselinelevel of
the criteriondrinkingmeasure,termsfor site main effects
andsiteby treatmenteffects,termsfor siteby matchingvari-
/JANUARY
1997
treatmentsthey were abstinentmore than 90% of the time
andat Month 15 therewasonly a slightdecrementin abstinence.Outpatientsubjectsaveragedslightlymoreabstinent
days per monthat baseline,but were abstinentmore than
80% of the daysat posttreatment,
with only a slightdecrementat the 15-monthfollow-up.
Survival analysiswas usedto examineelapsedtime to
first drink and to first heavy-drinkingperiod(3 consecutive
daysof heavydrinkingdefinedas ->6 drinksperdayfor men
and ->4 drinksperday for women)for subjectsin botharms
of the study(Figure2). In the aftercarearm, approximately
35% of subjectsreportedcontinuedcompleteabstinence
throughoutthe 12 follow-up months;65% slippedor relapsedduringthat period.Analyzingthe time to threeconsecutiveheavy-drinking
days,whichis a measureof regular
drinking as opposedto a slip or lapse, 40% of aftercare
clientsreachedthat level of drinkingduringthe follow-up
period; 60% never had three consecutiveheavy-drinking
days.For the outpatientsubjects,19% maintainedcomplete
abstinence
throughout
thefollow-upandapproximately
46%
hada heavy-drinking
periodof threeconsecutive
daysby the
endof thefollow-upperiod.
able interactions, and interaction terms for both linear and
quadratictimefor eachof thesecovariates.Thesecovariates
adjustedfor anyinitialdifferences
andfor differences
attrib-
Main effectsfor typeof treatment
utable to site.
Aftercare arm. In an analysisadjustedfor only baseline
drinkingand sitedifferences,no significantmain effectsof
treatmentwereobserved.Estimatedmeansfor drinkingoutcomesare shownin Figure 1. When the sameanalysiswas
furtheradjustedfor the ten matchingattributesto adjustfor
all matchingeffects(notreflectedin Figure1 means),a small
but statisticallysignificanttreatmentby time effect (linear
p < .001) emerged:TSF clientsshowedslightlyhigherPDA
outcomes
(fewerdrinkingdays)towardtheendof follow-up.
No differencewasobservedin drinkingintensity(DDD). In
light of the presenceof CRU by treatmentinteractionsand
the unadjusted
patternshownin Figure 1, we concludethat
there were no clinically significantoutcomedifferences
amongthesethreeaftercaretreatments.
Outpatientarm. As reflectedin Figure 1, an analysiswith
baselinedrinkingandsitedifferencesascovariatesindicated
no statisticallysignificantbetween-treatment
differencefor
PDA and DDD. When adjustments
for matchingattributes
were added,there was a small but statisticallysignificant
treatmentby lineartime effectfor bothPDA (p • .001) and
DDD (p < .05) outcomes.
Thisreflecteda tendencyfor CBT
clientstohavehada slightlyhigherrateof drinkingdaysover
timethantheothertwo groups.However,in no singlemonth
was there a significantdifferenceamonggroups.Again, in
light of significantsite by treatmentinteractionsand the
smallabsolutemagnitudeandshiftingpatternof effects,we
concludethat therewere no consistentand clinically meaningful differencesin efficacyof thesethreetreatments.
Secondaryoutcomevariables.While thea priorihypothesistestsall arebasedonthetwoprimarydependent
variables,
There were threeindicatorsof potentialmatchingeffects
in the latentgrowthanalyses.The attributeby treatmentinteraction(ATI) indicatedwhethertherewas an interactionin
thehypothesized
directiononaverageovertheentirefollowupperiod(Months4 to 15).A significantATI effectprovided
evidenceof a matchingeffect. There were also two indicatorsof whethertheATI changedsignificantlyoverthecourse
of the posttreatmentperiod: (1) an ATI by linear time
(ATI x T1) effect, and (2) an ATI by quadratic time
(ATI x T2) effect. Analysesinvolving thesetime effects
werecenteredat themidpointof thefollow-upperiodandindicatedwhetherthe ATI was shiftingin a linear (T1) or
curvilinear(quadratic)fashion(T2). Since time contrasts
were nondirectional,significantinteractionsrelatedto time
weretestedona monthby monthbasisto determinehowthey
werechangingovertime.
Results
Drinkingfrom baselinetofollow-up
Thereweresubstantial
positivechangesin PDA andDDD
for both aftercareand outpatientsubjectsfrom baselineto
eachof the follow-upmonthsas shownin Figure 1. These
improvementswere sustainedduringthe follow-up period
with only slightdeteriorationat 1-yearposttreatment.
Prior
to entryinto their inpatientor day hospitaltreatment,aftercare subjectswere abstinentaround20% of the daysper
month.In themonthimmediatelyfollowingProjectMATCH
PROJECT
MATCH
Percent of Days Abstinent by Treatment:
Aftercare
Arm
100
RESEARCH
GROUP
15
Percent of Days Abstinent by Treatment:
Outpatient Arm
100
80
8O
----
CBT
-.-
MET
........ TSF
60
40
20
20. •
0
0
Base
Treat
4-6
line ment
7-9
10-12
Base
13-15
Aftercare
Arm
20
4-6
7-9
10-12
13-15
Follow-up
Time of Measurement(Months)
Time of Measurement(Months)
Drinks per Drinking Day by Treatment:
Treat
line merit
Follow-up
Drinks per Drinking Day by Treatment:
Outpatient Arm
20
&
',/
• 10
=
5
Base
Treat
line merit
4-6
7-9
10-12 13-15
Follow-up
Timeof Measurement
(Months)
Base
Treat
line rnent
4-6
7-9
10-12 13-15
Follow-up
Timeof Measurement
(Month•)
FIGURE
1. MonthlyPDA andDDD outcomes
forbaseline
(averaged
over3 months
priortotreatment)
andforeachmonthof theposttreatment
period(months
4-15) for outpatient
andaftercare
arms.If a personwasabstinent
duringa givenmonth,theDDD scorewaszero.
PDA and DDD, thesemeasuresare not the sole meansof as-
of the dependent
measurebeinganalyzedandalsoadjusted
sessingoutcome.A rangeof otheroutcomevariableshave
been assessed,
includingother drinking-related
measures,
for the main effects of CRU and for the CRUX
use of substances other than alcohol and measures of social
andpsychological
functioning.
Thesediversemeasures
provide a fullerpictureof themaineffectsof thethreeProject
MATCH
treatments. Some of these variables will also be
usedin testingsomespecificmatching
hypotheses,
butthose
analysesarebeyondthe scopeof thepresentarticle.
Analysesof selectedsecondaryoutcomemeasuresinvolved two 3-monthtime periods:Months7-9 and 13-15.
Theseperiodswere chosenbecausemanyof the outcome
measureswere assessed
duringthe in-personinterviewsat
Months9 and 15. Continuousoutcomevariables(seeTable
3) wereanalyzedby repeated-measures
analysisof variance.
In theseanalysesof variance,we covariedthebaselinevalue
Treatment
Interaction.
We didnotattemptto adjustfor matchingeffects
becauseof thecomplexityof suchanalyses.
Two discrete outcome variables also were evaluated: com-
positeoutcomeandotherdruguse.The compositemeasure
of outcome,describedin Zweben and Cisler (in press),has
four levels that combineinformationaboutdrinkingand
drinkingconsequences
to yield a categoricalmeasureof outcome: 1 = no drinking,2 = moderatedrinkingand nonrecurrentproblems,3 = heavydrinkingor recurrentproblems,
and 4 = both heavy drinkingand recurrentproblems.The
composite
measure
takesintoaccounteventshappening
duringthemostrecent3 months.It doesnottakeintoaccountdurationof drinkingepisodes.
Forthismeasure,
weanalyzedthe
treatmentmain effectsusing log-linearmethods(Bishop
16
JOURNAL
OF STUDIES
ON ALCOHOL
/ JANUARY
1997
Three of the secondaryoutcomemeasuresshowedtreatmenteffectsthatachievedBonferroni-corrected
significance
levels:drinkingconsequences,
the compositeoutcomeand
0.8.
the time-to-event
measures. There was a treatment main ef-
fectfor drinkingconsequences
(assessed
usingtheDrlnC) in
the outpatientarm. Using the Duncanmultiplerangetest,
TSF clientswereshowntohavefewerdrinkingconsequences
than clients in the other two treatments at Month 9, but at
0.2'
Month 15 thethreetreatments
did notdiffer significantly.
In the log-linearanalysisof the compositemeasure,the
only predictorswere treatmentgroupand CRU. In the outpatientarm, treatmentmain effectson the compositemeasureat Month 9 achieveda significancelevel of p = .0349,
which did not meet the Bonferronicorrectionfor signifi-
o
Time to Firit Drlnlc Deys
1
cance. At Month 15, however, treatment main effects did
.•
0.8'"'.............
0.$
achieve a Bonferroni-correctedsignificance level of
outpMient p = .0024. As shownin Table 4, therewas,for example,a
higherpercentage
of TSF clientsin Category1 (nodrinking),
relative to the percentages
for the CBT and MET clients.
Therewereno statistically
significanttreatmentmaineffects
in thecomposite
outcomedatafor theaftercarearm clients.
In analyzingthe time-to-eventoutcomes,no treatment
l....Aftercare
main effects were found in the aftercare arm. There were,
Timeto 3 ConsecutiveHeavyDrinkingDays:Days
FIGURE2. Time to event (survival)curvesfor outpatientand aftercare
groupsduringthe 12-month
posttreatment
periodfortimeto firstdrinkand
timeto 3 consecutive
drinkingdays
however,significant
effectsamongtheoutpatient
clients.For
outpatients,
for time to first drink,the proportionalhazards
analysisyieldedap valuefor treatment
maineffectsof.0001;
there were also statisticallysignificantCRU main effects
(p = .0007)andCRU by treatmentinteractions
(p = .0065).
TSF clients had the best outcome on this measure, with 24%
et al., 1975).For thedruguseoutcomevariable,we usedlogisticregression.
Becauseof the relativesparsityof useof
mostillegalsubstances
otherthanmarijuana,we collapsed
anyuseof anyillegalsubstance
intoa binaryoutcomemeasure.
Finally,wealsoconducted
testsfor treatment
maineffects
usingthe two time-to-first-event
measures
reportedearlier:
time to first drinkandtime to first episodeof threeconsecutive daysof heavydrinking.Thesetestsinvolvedproportionalhazardsanalyses(Cox, 1972).
Analysesof the continuoussecondaryoutcomemeasures
arereportedin Table 3, anddatafor the composite
variable
arein Table4. We haveapplieda Bonferronicorrectionfor
testingthesenine secondary
outcomemeasures;
in the far
right columnof Table 3, an asteriskindicatestreatment
effects are significantin a nondirectionaltest at alpha--
avoidingany drinking in Months 4-15, while the corresponding
figuresfor CBT andMET were 15%and 14%,respectively.
Whenwe analyzedthemorestringent
criterionof
threesuccessive
daysof heavydrinking,treatmentmaineffectsweresignificantin theoutpatientarm only,p = .0016;
CRU maineffects(p = .0054) andCRU by treatmentinteractions(p = .0127) alsowerepresent.Onceagain,the TSF
conditionhadthebetteroutcome,with 53% not reachingthe
criterion,followedby MET with 49% andCBT with 48%.
Matchingoutcomes:Primary hypotheses
Testsof the primarymatchinghypotheses
over the 4- to
15-monthfollow-upperioddemonstrated
few matchingeffects.A summaryof thesignificant
resultsof the 16contrasts
of theprimarymatchinghypotheses
appearsin Table5.
Aftercarearm:Percentdaysabstinent
(PDA). Ignoringef-
.05/9 = .0056. Not shown in the tables are the results for
fects over time, there were no Bonferroni-correctedclient at-
druguseandthe time-to-event
analyses.
We did notdetect
any statisticallysignificanttreatmentmain effectson drag
tributeby treatment
interactions
(ATI) forPDA for anyof the
primaryhypotheses
in theaftercarearm.Whenattributeby
treatmentby time effectswereexamined,only one signifi-
useoutcomein either studyarm. In the outpatientarm, the
rateof anyuseof anyillegalsubstance
duringthepreceding
90 dayswas30% of clientsat Month9 and29% at Month 15;
themajorityof substance
usewasmarijuana.In theaftercare
arm, the ratesat Months 9 and 15, respectively,were 18%
and 19%.
cant interactionwas found: the meaning seekingclient at-
tributeby treatment(TSF vs CBT andMET) by lineartime
(p = .01). Duringthelatterhalf of theposttreatment
period,
thoseclientstreatedin TSF whohadhighermeaningseeking
were more likely to have proportionately
more abstinent
PROJECT MATCH
RESEARCH
GROUP
17
TABLE3. Treatmentmaineffectsoncontinuous
secondary
outcomemeasures
Treatmentgroup
CBT
Variable
Mean(-+SD)
MET
Na
Mean(-+SD)
AFTERCARE
TSF
Na
Mean(-SD)
Na
Trtmteffects
0
STUDY
Drinkingconsequences
Baseline
Month 9
Month 15
59.3 +--23.5
19.6 -+ 27.9
19.3 _ 29.3
164
57.4 _+22.1
20.0 _+26.8
16.9 _+ 23.1
168
60.7 -+ 23.3
19.4 ___
28.3
21.2 -+ 29.0
161
.9690 •
102.4 +__119.0
77.7 - 106.1
81.0 _+ 109.2
103
72.9 +__86.9
70.0 -+ 100.5
58.0 - 80.6
93
93.8 -+ 116.6
74.2 _+96.8
77.2 - 101.4
88
.7194
35.9 --+32.6
42.6 _+ 31.5
43.7 __+35.7
166
40.2 ___33.1
42.4 _ 30.9
46.0 _ 35.8
138
40.0 _+ 33.8
39.6 - 33.2
43.9 - 38.1
159
.5427
3.02 _+0.62
3.37 -+ 0.53
3.37 -+ 0.52
209
3.05 --+0.57
3.36 -+ 0.49
3.40 __+
0.47
209
2.95 - 0.64
3.33 _+0.50
3.31 -+ 0.51
178
.4555
10.36 +__8.56
7.76 -+ 8.16
7.95 __+8.93
205
10.04 __+8.62
8.45 _ 8.03
8.77 +__8.56
197
11.41 __+
9.13
8.93 _+ 8.75
8.75 _+ 8.84
179
.2929
0.23 + 0.21
0.17 +__
0.20
0.15 + 0.20
227
0.23 ___
0.21
0.14 +__
0.19
0.16 +__
0.21
221
0.23 __+
0.22
0.16 _+0.19
0.15 _ 0.19
199
.5938
GGT
Baseline
Month 9
Month 15
Percentdayspaid work
Baseline
Month 9
Month 15
Social Behavior Scale
Baseline
Month 9
Month 15
BeckDepression
Inventory
Baseline
Month 9
Month 15
ASI Psych.Severity
Baseline
Month 9
Month 15
OUTPATIENT
STUDY
Drinkingconsequences
Baseline
Month 9
Month 15
44.6 +__
21.2
21.4 _+24.3
19.7 _ 23.1
201
46.2 __+
21.8
23.5 _ 23.2
19.9 __+
23.4
202
45.7 _ 22.6
16.7 -+ 21.8
15.9 _+20.7
245
.0045*
82.7 -+ 93.0
65.8 -+ 74.8
71.8 _ 87.3
224
78.4 _ 90.9
66.3 _+ 81.6
67.8 _+ 82.8
206
72.1 -_+88.6
61.1 __+
76.2
61.7 -+ 75.3
240
.7610
46.8 _+31.3
45.8 _ 31.5
48.4 _+32.6
178
44.4 _+ 33.0
49.6 __+
30.9
54.1 _+ 31.7
200
49.8 _+ 30.7
47.1 _+ 31.6
52.0 _+33.0
183
.2342
273
.9041
256
.3020
274
.3202
GGT
Baseline
Month 9
Month 15
Percentdayspaid work
Baseline
Month 9
Month 15
Social Behavior Scale
Baseline
3.23 -_+0.51
Month 9
Month 15
3.40 __+
0.46
3.44 __+
0.45
3.18 +__
0.50
253
3.37 +--0.47
3.44 - 0.46
3.24 _+0.49
245
3.43 _+0.45
3.44 __+
0.46
BeckDepression
Inventory
Baseline
10.30 -+ 8.28
Month 9
8.11 _+7.99
Month 15
7.32 _+ 7.85
9.54 ___7.37
233
7.02 _+7.41
10.06 +__8.21
234
6.56 --+6.78
6.80 _+6.96
7.08 -+ 7.84
ASI Psych.Severity
Baseline
Month 9
Month 15
0.21 __+
0.20
0.13 _+0.19
0.12 _+0.19
253
0.19 _+0.19
0.11 - 0.17
0.11 --_0.16
249
0.20 ___
0.19
0.12 ___
0.16
0.10 _+0.15
Note:Months9 and15 referto thepreceding
90-dayperiod.
•Thedatain thetableareforthosesubjects
whohadnonmissing
valuesatall threetimepoints,
hencethesample
sizeisthesame
for all time pointswithin eachscale.
bThetreatmenteffectscolumncontains
p valuesfor a nondirectional
2 df testfor maineffectsof treatment.Thoseeffectswhose
significance
exceeds
theBonferroni-adjusted
levelof 0.0056aremarked
withanasterisk.
Treatment
bytimeinteractions,
if any,
are indicated with footnotes.
cThereis a significant
timeby treatment
effectfor drinkingconsequences
(DrInC),p = .0466.By theDuncantest,however,
the
treatment
groupsdonotappearto differsignificantly
at eithertimepoint.
18
JOURNAL
OF STUDIES
ON ALCOHOL
TABLE4. Treatmentmaineffectsfor composite
outcomevariable
1
2
3
4
N
AF'rERCARE STUDY - MONTH 9
CBT
MET
42.6
44.8
13.2
6.6
15.2
19.0
26.8
31.8
250
242
TSF
Combined
44.8
44.1
8.6
9.5
16.4
16.9
30.2
29.6
232
724
AF'rERCARE
STUDY
CBT
MET
TSF
Combined
48.0
42.5
47.3
45.9
7.0
8.3
6.6
7.3
34.0
35.4
34.1
34.5
244
240
226
710
43.8
288
41.6
35.9
40.3
296
320
904
41.0
43.3
37.8
40.6
283
284
312
879
OUTPATIENT
20.8
17.4
MET
TSF
Combined
23.3
31.9
25.6
14.5
14.4
15.4
CBT
MET
TSF
Combined
24.7
30.3
35.6
30.4
15
11.1
13.8
11.9
12.3
STUDY
CBT
OUTPATIENT
- MONTH
- MONTH
STUDY
- MONTH
14.1
14.1
9.3
12.4
20.1
12.3
17.3
16.6
creased over time.
9
18.1
20.6
17.8
18.8
1997
meaningseekingcontrasts
indicatedchanges
overtimein the
directionof the specifiedcontrast,althoughnoneof the valuesreachedthe .05 level of significance.
Aftercare arm: Drinks per drinking day (DDD). Again
therewere no significantATI effectsfor the DDD outcome
in the aftercarearm. However,therewas a significantinteraction effect for typologyby treatmentby time indicating
that the contrastbetweenType B (more severe)subjects
treatedwith CBT and TSF versusType B subjectstreated
with MET shiftedover time (linear, p < .05). However, no
singlemonthlycontrastreachedthe .05 levelandp valuesin-
Compositeoutcomecategory(%)
Treatmentgroup
/ JANUARY
15
Notes:The compositeoutcomecategories
are as follows: 1 = no drinking
duringthe periodof assessment;
2 = moderatedrinkingand nonrecurrent
problems;3 = heavydrinkingor recurrentproblems;and4 = bothheavy
drinkingandrecurrentproblems.The composite
measuretakesintoaccount
eventshappening
duringthemostrecent3 months.It doesnottakeintoaccountdurationof drinkingepisodes.
daysthanthosetreatedin CBT or MET. When attributeby
treatmentby time effectswere found,follow-up analyses
wereperformedto examinewhetherthe hypothesized
contrastsimplychangedovertime or producedsignificantdifferencesduringanyof thefollow-upmonths.The outcomes
in termsof p valuesof monthby monthtestsof the specific
Thus,in the aftercareconditiontherewas no unequivocal
supportfor anyof thematchinghypotheses
for eitherPDA or
DDD outcomes
overthe12-monthfollow-upperiod.Although
thereweretwo significant
clientattributeby treatment
by time
effectsduringtheposttreatment
period,neitherwasstatistically
significantfor any singleposttreatment
month.However,the
meaningseekingcontrast
didhavemonthswherethecontrasts
weresignificant
(p < .05)in ananalysis
withfewercovariates.
Outpatientarm: Percentdaysabstinent.Therewasonesignificantclientattributeby treatmentinteraction(ATI) for the
PDA drinkingoutcome
in theoutpatient
armforoneoftheproposedcontrasts
ofthepsychiatric
severity
hypothesis
(CBT vs
TSF, p = .01).It indicatedthatthelesstheclient'spsychiatric
severityscore,thegreaterhis/herpercentdaysabstinent
when
treatedwith TSF, comparedto CBT. No othermatchinghypothesis
contrast
demonstrated
a significant
overallATI.
This samecontrast(CBT vs TSF) of thepsychiatricsever-
ity matchinghypothesisalso demonstrated
a significant
client attributeby treatmentby time interaction(quadratic
time,p < .05). Examinationof this contrastby monthindicatedthat there were significant(p < .05) effectsdemonstratedfrom Months5 through11 on PDA. Figure 3 shows
TABLE5. Treatmentmatchinganalyses:
Summaryof significanteffectsfor specific treatment contrasts
Matching
variable
Treatment
contrasts
Typeof
effect
Outcome
variable
Months
p < .05
OUTPATIENT
Psychiatric
severity
Motivation
Conceptual
CBT vsTSF
MET vsCBT
TSF vsMET
ATI
ATI by time
ATI by time
ATI by time
PDA
5-11
PDA
DDD
15
None
ATI by time
PDA
None•
ATI by time
DDD
None
level
AFTERCARE
Meaning
seeking
Typology
TSF vs
CBT andMET
MET vs CBT
Notes:Thistablesummarizes
resultsof latentgrowthanalyses
covaryingbaseline
drinkingvaluesaswell asCRU andCRU by treatment
effectsovertime.Typeof
effectindicateswhethertheattributeby treatmentinteraction(ATI) and/ortheATI
over time were significant.The monthscolumnsummarizes
resultsof monthly
contrasts.
aMonthlycontrasts
formonths5 through14werep <.05 withonlybaseline
drinkingcovaried,andmonthlycontrasts
for months11, 12 and 13werep <. 10 when
adjustedfor baselinedrinking,CRU andCRU by timeeffects.
PROJECT MATCH
RESEARCH
GROUP
19
PsychiatricSeverity
Month 4
0.0
0.1
0.2
0.3
Month
0.0
0.1
0.2
Month $
0.4
0.5
0.0
0.1
7
0.3
0.4
0.5
0.0
0.1
Month 10
0.0
0.1
0.2
0.3
0.1
0.2
0.3
0.3
Month
8
0.2
0.3
0.4
0.5
0.0
0.1
0.4
0.5
0.0
0.1
0.2
0.3
0.4
0.5
0.0
0.1
0.5
0.0
0.1
0.2
0.3
0.3
0.4
0.5
0.2
0.3
0.4
0.5
0.4
0.5
0.4
0.5
Month 12
0.4
0.5
0.0
0.1
Month 14
0.4
0.2
Month
Month 11
Month 13
0.0
0.2
Month 6
0.2
0.3
Month 15
0.4
0.5
0.0
0.1
0.2
0.3
FIGURE
3. Monthlyposttreatment
plotsof percent
daysabstinent
fortreatment
bytimeby attribute
interaction
forpsychiatric
severity
contrast
between
CBT
andTSFamong
outpatients.
Theinteractions
atmonths
5-11weresignificant
in thepredicted
direction
(p's< .05).Thevertical
axisrepresents
percent
days
abstinent
andthehorizontal
axisrepresents
psychiatric
severityscores.
20
JOURNAL
OF STUDIES
ON ALCOHOL
the interactioneffectsfor eachmonthof the follow-up. In
Months5 through11,theregression
linesintersectat a value
of approximately0.4 on the AddictionSeverityIndex (ASI)
psychiatriccompositescore.A posthoc analysisof covarianceconductedon a subsample
of participantswith an ASI
psychiatricscoreof zero revealedsignificantlybetterPDA
outcomes in the TSF condition than in the CBT condition.
Similaranalyseswereunableto demonstrate
a significantadvantageof CBT over TSF in high psychiatricseverityparticipants,regardless
of theuseof differentsubsamples
based
on progressivelyhigher ASI psychiatricseverity cutoff
scores.AlthoughslopelinesindicatedthatCBT-treatedparticipantswereexceedingTSF participants
at higherlevelsof
theASI scale,a clearconclusion
cannotbe drawnregarding
whethertherearebetteroutcomes
for highpsychopathology
participantstreated with CBT. It is thus concludedthat
clientswithoutpsychopathology
hadmoreabstinentdaysif
treatedwith TSF ratherthan CBT, but this TSF advantage
disappeared
aspsychopathology
increased.
Also in the outpatientarm, the motivationhypothesis,
whichstatedthatsubjectslowerin motivationwoulddobetter in MET thanin CBT, demonstrated
a significantATI by
time interaction(linear,p < .01). However,thecontrastwas
significant
(p < .05) at only 1month(Month15)for thePDA
outcome.As shownin Figure 4, the relationshipbetween
CBT andMET treatments
for thelessmotivatedsubjects
beganwith thelessmotivatedsubjects
initiallydoingbetterin
CBT comparedto MET, butthiseffectreversedovertime so
that by the end of follow-up, the less motivated subjects
treatedin MET had a greaterpercentage
of abstinentdays
comparedwith the CBT clients.Therewaslittle difference
betweentreatmentsover the follow-upperiodfor subjects
with highmotivationto change.
Outpatientarm: Drinks per drinkingday. There were no
significantATI effectsfor any of theprimaryhypotheses
in
DDD. Therewasa significant
clientattributeby treatment
by
time effect for the conceptuallevel hypothesis(quadratic,
p < .01) for DDD which indicatedthat the relationshipof
conceptual
level andtheMET vs TSF contrastshiftedover
time.However,noneof themonthlytestsof thehypothesized
contrastapproached
a .05 levelof significance.
In fact,in the
last monthof follow-upthep valueindicateda significant
contrastoppositeto thathypothesized.
In summary,in the outpatientarm of the trial therewasa
matchingeffectfor one specifiedcontrastof thepsychiatric
severityhypothesis.
Althoughthe originalconceptualization
of thishypothesis
wasthatindividualshighin psychopathologywouldhavebetterdrinkingoutcomes
withCBTratherthan
TSF, resultsindicatedthat there was no reliable differencein
theoutcomes
of highpsychopathology
subjects.
On theother
hand, subjectswithout psychopathology
had significantly
moreabstinence
in 7 of the 12follow-upmonthswhentreated
withTSF ratherthanCBT. The TSF advantage
overCBT was
onaverageapproximately
4 moreabstinent
dayspermonth.
/ JANUARY
1997
In additiontothelatentgrowthanalyses
of treatment
matching effects,traditionalrepeatedmeasures
MANOVA
analyses
were alsoconducted
for eachof the primarymatchinghypotheses.
Althoughtherearesomedifferencesin thesetwoanalyticapproaches,
theresultsof thesemoretraditionalanalyses
wereconsistent
with the majorfindingsof the latentgrowth
analyses,
butgenerallywithp valuesgreaterandthuslesssignificantthanthosefoundwiththelatentgrowthanalyses.
Treatmentsitedifferencesanalysis
Main effectsfor treatmentsitewerepresentin botharms
for PDA outcomes,but only in the outpatientarm for the
DDD outcomes.
By design,treatmentsitesrepresented
differentialclientheterogeneity
with theiruniquecontributions
to thepoolof subjects.
Treatmentsiteeffectsmay be dueto
thesedifferencesin the clientpopulationsor environmental
factorsnotmeasured
by covariates.
Suchdifferencesshould
havelittle effecton testsfor client-treatment
matching,however, sincethe covariateset usedin our analysesincluded
termsfor bothsiteand siteby time interactions.
There were several site by treatmentinteractionsthat
couldpotentiallyhavebeendue to variationsin the implementationof treatmentacrosssitesand,as such,couldrepresent a potential threat to the validity of the tests for
client-treatment
matching.However,treatmentprocessdata
indicatedthat treatmentimplementation
was relativelyuniformacrosssites.Thatis,therewereno substantial,
clinically
meaningfuldifferences
with respectto treatmentimplementation, perceivedtherapeuticalliance, sessiontype (emergency,collateral),and severalother variablesacrosssites
thatmightbe expectedto affect drinkingoutcomes.Site by
sitetestsof thematchinghypotheses
indicatedthattheoverall matchingresultsaregeneralizable
acrosssites.
Drinking outcomeeffectsfor clientattributes
Sincetheclientattributes
chosen
for thematching
hypotheses
canaffectoutcomes
independent
of treatmentcondition,a separateanalysiswasconducted
to examineclient
attribute effects on both PDA and DDD
outcomes. Since at-
tributesconstituted
a relativelylargepool of variablesand
werecorrelatedwith oneanotherto varyingdegreesranging
from a Pearsonfirst order correlationof .00 to _+.50,a back-
ward eliminationapproachwas usedin orderfirst to eliminatenonsignificant
effectsandthento examineeffectsof the
retainedvariables.Table 6 reportsthe significantmain and
time dependentinteractioneffects.
For aftercaresubjects,only genderpredictedthe percent
daysabstinentover the entirefollow-upperiod,with male
subjects
havingfewerabstinent
days.Althoughtherewasno
maineffectof psychiatricseverityon outcome,thisattribute
did interactwith time to predictPDA outcome.Towardthe
end of the follow-up period subjectshigher in psychiatric
PROJECT MATCH RESEARCH GROUP
21
Motivation
Month
Month $
4
Month 6
t(•
2
4
6
8
10
12
14
1(•
::
2
4
6
8
10
12
14
:
Month 8
Month 9
Month 10
Month 11
Month 12
Month
Month
Month
Month 7
10o
10o
2
4
6
8
10
12
14
13
14
16
lOO
2
4
•
•
10
12
14
2
4
6
8
10
12
14
2
4
6
8
10
12
14
FIGURE
4. Monthly
posttreatment
plotsofpercent
daysabstinent
fortreatment
bytimebyattribute
interaction
formotivation
contrast
between
CBTandMET
among
outpatients.
Theinteraction
atmonth
15wassignificant
inthepredicted
direction
(p< .05).Thevertical
axisrepresents
percent
daysabstinent
andthe
horizontalaxisrepresents
motivationscores.
22
JOURNAL
OF STUDIES
ON ALCOHOL
TABLE6. Significantmaineffectsandtime effectson PDA andDDD out-
/ JANUARY
1997
Discussion
comes for client attributes
Attributes
Aftercarep values
Outpatientp values
PDA
PDA
DDD
<.001
<.001
Alcohol involvement
Main
Gender
Main
DDD
.035
.017
Linear time
Motivation
Main
Psychiatricseverity
Linear time
Quadratictime
Supportfor drinking
Main
.012
.015
.022
.010
.005
.026
Sociopathy
Linear time
<.001
.029
Typology
Quadratictime
Note:p valuesrepresent
the resultsof a stepwiseproceduredoneacrossall
CRUs with CRU andCRU by treatmentinteractiontermsin the modelin
whichinitiallyall the matchingvariableswereincluded.Theseresultsrepresentthesignificantmainandtimeeffectsof thevariablesretainedafterthe
laststepof backwardselimination.
severityhadfewerdaysabstinent
comparedto thoselowerin
psychiatricseverity.
Client attributesdemonstratedgreaterinfluenceon the
numberof drinksperdrinkingday(DDD) oncea subjectbegan drinking.Higher alcoholinvolvement,beingmale and
havingmoresocialsupport
for drinkingwereeachassociated
with moreDDD duringfollow-up.In addition,the prognostic effectsof gender(male)andpsychiatricseverity(greater)
on DDD were more pronouncedas time increasedin the
follow-upperiod.
For the outpatientsubjects,fewer clientattributepredictorsweresignificant.The moremotivatedthe subjectwasat
intakeandthe lessthe socialsupportfor drinking,thebetter
werethedrinkingoutcomesin termsof bothPDA andDDD.
Level of sociopathyinteractedwith time as a predictorof
outcome:greatersociopathy
wasassociated
with worseoutcomesearlyin thefollow-upperiodbutnot later.
ProjectMATCH was designedas a prospectivestudy,
with tenprimarymatchinghypotheses
(containing16 contrasts)testedwith two dependent
variableschosena priorito
representtreatmentoutcome.Within theseconstraints,
we
foundvery limitedevidencefor eithermain or matchingeffectsfor thethreetreatments
studied.It isplausible,however,
thatinformativeeffectsmay be foundasotheroutcomevariablesare examinedand as secondarymatchinghypotheses
are tested.This reporthasfocusedon themeasures
andprimaryhypotheses
chosena prioriby theProjectMATCH ResearchGroupto constitutethe main trial. Similarcare will
now be devotedto conductingand subsequently
reporting
analyses
for a priorisecondary
matchinghypotheses
andfor
secondary
outcomevariableswith all hypotheses.
Tests of the specificmatchinghypothesesevaluatedin
ProjectMATCH providedlimitedsupportfor thegenerichypothesisthat client attributeswould interactwith treatment
modalityto differentiallyaffectdrinkingoutcomes.
Onlyone
clientattributeexaminedhadan overallmatchingeffectthat
wasnot time dependent.
Outpatients
withoutpsychopathology had significantlymore abstinencewhen treated in
Twelve-StepFacilitation(TSF) thanthosetreatedin Cognitive BehavioralCoping Skills Therapy (CBT), but as psychiatricseverityincreased,the TSF advantageover CBT
disappeared.
Sincethe outpatientsamplehadfewer participantsat thehighendof psychiatricseverity,it wasnotpossible to evaluate completely whether CBT led to
significantlymoreabstinentdaysthanTSF at the high end
of severity. Overall, the outpatientsample was slightly
lower on the ASI psychiatric composite score (mean
[_ SD] =. 19 +__.
19) thanotheralcoholismtreatmentsamples
(e.g.,McLellan et al., 1992;mean= .24 _+.22). Individuals
with currentsuiciderisk, homiciderisk or acutepsychosis
wereexcludedfrom thepresenttrial.
Thesepsychiatricseveritymatchingresultssuggestthat
thereis someadvantage
to assigning
outpatient
clientswithout psychopathology
to TSF treatment.The largestdifferenceoccurredat Month 9, whenmatched(TSF) participants
hadapproximately
87% daysabstinentversus73% daysabstinentfor mismatched
(CBT) participants.
Definitiveclienttreatmentmatchingrecommendations
for outpatientclients
with moderateto high psychiatricseveritycannotbe made
based on Project MATCH results.Since no psychiatric
severitymatchingeffectswerefoundin the aftercarestudy,
noclient-treatment
matchingrecommendations
canbe made
for the aftercaresetting.
One otherclient attributeamongoutpatients,
motivation,
interacted
withtreatment
modalitiesashypothesized,
butthis
interactioneffectchangedovertime anddemonstrated
a significantdifferenceduringonlythelastmonthof thefollow-up
period.Otherclientattributes,
meaningseeking,conceptual
levelandtypology,werealsoobserved
to havematchingcontrasteffectsthatchangedovertime, but at no time pointdid
the hypothesized
contrastreachsignificance.Evidencefor
eachof theseeffectsoccurredin onlyonearmof thestudy.
Aside from psychiatricseverity,the mostnotablematching findingsinvolvedmeaningseekingand motivation.In
aftercare,clientshigherin meaningseeking(i.e., thosewho
at intakeevidencedlesspurposein life andaspiredto experiencegreatermeaning)weresomewhatmoreresponsive(in
termsof PDA) to TSF thanto othertreatments.This pattern,
which was modest and consistent in direction across sites,
hadbeenpredictedbecausethestrongtwelve-stepemphasis
on spiritualitywas hypothesized
to appealparticularlyto
clientsseekinggreatermeaningin life. Evidencefor thisinteractionwas lackingduringthe first 6 monthsafter treatment,emergingonly in thelatterhalf of thefollow-upyear.
PROJECT
MATCH
Outpatientclientslow in motivationultimatelydid better
in MotivationalEnhancement
Therapy(MET). At thebeginningof theposttreatment
period,however,CBT appearedto
be superiorto MET in PDA for clients less motivatedto
change.Over the courseof the follow-up,the outcomesfor
thetwo treatmentsreversed,with MET becomingsuperiorto
CBT, indicatinga possibledelayedeffect.DDD outcomes
are
consistent
withthisfindingbutnotstatistically
significant.
An ongoing3-yearfollow-upstudyof outpatientsubjects
shouldyieldadditionaldatathatwill shednewlightoneffects
thatshiftor emergeovertime. Furthermore,
theplannedexaminationof thecausalchainsthatwereproposed
for eachhypothesisshould reveal whether hypothesizedmediating
mechanisms
operatedas assumed,and may offer explanationsforthepresence
orabsence
of hypothesized
interactions.
In summary,thislarge-scale,
randomized,
clinicaltrial has
detectedsimple(i.e., non-time-dependent)
matchingeffects
in the directionspredictedfor only one of the ten client attributeshypothesized
to interactwith the chosentreatment
modalities.Exceptfor psychiatricseverity,thereis not convincingevidenceof major treatmentmatchingeffects.Observedeffectsare sufficientlysmallandcircumscribed
that,
againwith theexceptionof psychiatric
severity,we canconclude that they are clinically insignificantwhen making
triaging decisionsto individualtherapyemployingthese
threetreatments.
Matchingclientswith the identifiedattributes to thesetreatmentmodalitiesdid not appreciablyenhance treatmenteffectivenesson our primary drinking
outcome
measures.
Psychiatricseverityasa matchingattributedeserves
more
intensiveexaminationbecause,in the outpatientstudy, it
aloneinteractedwith treatmentto affectdrinkingacrossmost
of the 1-yearfollow-upperiod.A numberof otheralcoholism
treatmentstudieshave also found significantpsychiatric
severitymatchingeffects(Cooneyet al., 1991;Kaddenet al.,
1989;McLellan et al., submittedfor publication,1983a).Project MATCH, however,is the first studyto examinepsychopathologyby treatment interactionswith a 12-step
approachamongthe treatments
examined.The findingof a
TSF advantageover CBT in individuals without psychopathology,
but not in individualswith moderateto high
psychopathology,
suggests
thatprocessanalysesshouldlook
for someingredients
in theTSF conditionthataredisrupted
by
psychopathology.
The ASI psychiatriccompositescoreis a globalmeasure
thatcombinessymptoms
of anxiety,affective,psychoticand
personalitydisorders.Furtheranalyseswill examinehow
well more specific, diagnostic-based
measuresof psychopathology
performasmatchingvariables.
Althoughit isneverpossible
toprovethenullhypothesis,
the
powerof thepresentstudyto detectmatchingeffects,andits
careful,rigorousimplementation,
makethelackof substantial
findingsparticularlynotable.Ourdataprovidelittleevidence
to supportwidelyheldviewsregardingthepotentialvalueof
matchingclients,at leaston thebasisof nineof the clientat-
RESEARCH
GROUP
23
tributestested,to any of the treatments
offeredasindividual
therapyin thisstudy.Theseresultssupport
widerlatitudein the
triagingprocess
withlessneedtomatchbasicclientcharacteristicsto anyof thesethreetreatments,
if theyareimplemented
carefullyasindividualtherapyby well-trainedtherapists.
It shouldbe underscored,
however,thatthelackof support
for matchinghypothesesinvolvingthese three particular
treatmentsdoesnot addresspotentialmatchingeffectsthat
possiblycouldappearif morediversetreatmentdeliverysystemswerecontrasted
(e.g.,inpatientvsoutpatient
treatments,
groupvs individualtherapies,socialsystemtherapies[such
as the communityreinforcementapproachor behavioral
maritaltherapy]vs individualtherapies,or pharmacological
therapiesvs psychosocial
therapies).Nor do thesefindings
holdfor all typesof substance
abuserswith varyingor multiplesubstances
of abuse,or thehomeless.
Althoughthesample gatheredfor this studywaslargeandheterogeneous,
it
doesnot fully representthe entirepopulationof alcoholdependentindividualsor othersubstance
abusers.
Finally,there
may be other client attributesor treatmentcontraststhat
could yield important matching information. Project
MATCH researchers
planto examinethe datasetfor potential matchinginteractions
in termsof additionalclientcharacteristics,differentcontrasts,secondaryoutcomevariables,
and more complextypesof matchinginvolvingcombinations of variables.
Althoughthe efficacyof the threetreatmentscannotbe
demonstrateddirectly sincethe trial did not includea notreatmentcontrolgroup,the strikingdifferences
in drinking
by clientsfrom pretreatment
levelsto all follow-uppoints
suggestthat participationin any of thesetreatmentswill be
associated
with substantial
and sustainedchangesin drinking. This is particularlytrue for the outpatientarm, where
the ProjectMATCH treatmentswere the only treatments
provided.One importantconclusionof thistrial is thatindividuallydeliveredpsychosocial
treatmentsembodyingvery
differenttreatmentphilosophies
appearto producecomparablygoodoutcomes
(HesterandMiller, 1995;Lambertand
Bergin,1994), a findinggenerallysupported
by evaluation
of a varietyof secondaryoutcomemeasuresas well. In fact,
the sustained,positiveimprovementfor clientsin all three
treatmentconditionsmay haveleft little roomfor matching
effectsto emerge.Implicationsare, of course,unknownfor
treatments
thatarenotmanualguided,notstructured
to produce and utilize a good therapeuticrelationship,or are
poorlydone.
Thetreatment
compliance
of theindividuals
in thistrialwas
high.Subjectsreceivedsubstantial
amountsof the specified
treatments.Complianceenhancement
procedures
(i.e., callingclientsbetweensessions,
sending
remindernotesandhavingcollateralcontacts)
andthegreaterattentionof individual
treatmentmay haveproduceda level of overallcompliance
that made it difficult for differences between treatments to
emerge.It ispossible
thatprevious
matching
studies
mayhave
reflectedvariationsin treatmentcompliance.
24
JOURNAL
OF STUDIES
ON ALCOHOL
Finally, researchfollow-up compliancealso was remarkable,reflectinganintensiveeffortonthepartof researchstaff
and paymentof clientsas an incentiveto returnfor followup. The overall effect of being a part of ProjectMATCH,
with its extensiveassessment,
attractivetreatmentsand aggressivefollow-up,mayhaveminimizednaturallyoccurring
variabilityamongtreatmentmodalitiesandmay, in part,ac-
/ JANUARY
1997
suchAA participation
is nevertheless
animportant
topicthat
will be addressed
morefully in a futurereport.
The performance
of MET relativeto CBT andTSF suggestedthat this four-session,
12-weektreatmentmodality
canbe usedin lieu of thesemoreintensiveambulatorytreatments, at least in the context in which it was delivered in this
suggesttheimportanceof examiningthepotentialimpactof
systemandtreatmentdeliveryattributesin additionto treatmentphilosophyandclinicallyeffectiveingredients.
Whereasparticipationin anyof thesethreetreatments
was
generallyassociated
with a sustained,goodoutcome,some
smallbut significantdifferencesamongthetreatmentson the
primaryoutcomemeasuresweredetectedin boththeoutpa-
trial.Thefactthatnootherhypothesized
clientattributemoderatedthe effectiveness
of this treatmentwith the rangeof
clientstreatedin this studysuggeststhat four sessionsof
MET may have more widespreadapplicabilitythanpreviouslythought.In fact, the reasonsomeof the matchinghypothesiscontrastsdid not receivesupportis that many of
them assumeda mismatchingeffect with the lessintensive
MET that did not materialize.An ongoingstudyof costeffectivenesswill examinewhetherMET may be a more
tient and aftercare arms. However, these effects were not ro-
cost-effective treatment than either CBT or TSF. However,
bustandrepresented
smallabsolutedifferencesin percentage
of daysabstinentor drinksperdrinkingday.In termsof other
outcomevariables,outpatient(but not aftercare)clientsin
we againcautionthattherewasno untreatedcontrolgroupin
thistrial andthatmanystepsweretakento ensurequalityof
treatment.In addition,theimpactof intenseandfrequentfollow-up effortsand corroborativecheckingof drinkingbehavior every 3 monthsmay also have affectedoutcomes.
Finally, MET subjectsalso attendedsome AA meetings.
Processdataandanalysisof sessionvideotapes
will enable
usto examinein greaterdetailthemechanisms
of actionfor
count for the favorable treatment outcomes. These outcomes
the TSF treatment
showed better outcomes on three mea-
suresstronglyinfluencedby continuous
abstinence(time to
first drink, time to first run of 3 heavy-drinkingdaysand
compositeoutcome),perhapsdue to a greateremphasison
abstinence in TSF relative to CBT and MET.
With the ex-
ceptionof an advantageto TSF in drinkingconsequences
at
Month 9, no othertreatmentdifferencesemergedon the remainingoutcomemeasuresat eitherMonth 9 or 15 for outpatientor aftercareclients.Thus,an overallimplicationfor
the field is that each of these three treatments can be used
with confidence,whenimplementedastheywerein theProject MATCH trial.
There are severalother interestingand importanttreatment-relatedimplications.Of particularimportanceis the
performance
of anindividuallydelivered12-step-based
treatment(TSF) andtheperformance
of thefour-session,
12-week
motivational-based
treatment(MET). AlthoughTSF mustbe
clearly distinguished
from AlcoholicsAnonymousand its
practices
andtraditions,
TSF is a 12-step-based
approach
that
encourages
AA attendance
andtheworkingof the 12 steps.In
particular,ProjectMATCH represents
thefirstdemonstration
in a randomizedclinical trial, controllingfor othertreatment
factors,of comparableoutcomesfrom a 12-step-based
approachandothertreatmentmethods.Onepotentiallimitation
in the useof TSF comparedto CBT may be thatit is perhaps
not aseffectivefor aftercareclientswith low meaningseeking.However,therearefew signsof theothermismatches
that
we had hypothesizedfor TSF (e.g., femalesand thosewith
higherconceptuallevelshavingpooreroutcomesin TSF). It
should be noted that AA attendance was not controlled in this
trial. Clientsin TSF did attendsignificantlymoreAA meetingsthandid thosein theothertwo treatments,
butclientsin
CBT andMET, particularlyin theaftercarestudy,wereoften
exposed(outsidethecontextof ourtreatment)toAA anda 12stepapproach.
While AA attendance
duringtreatmentdidnot
appearto be an importantmediatingvariablein this study,
MET, as well as for CBT and TSF.
CBT hashadanestablished,
research-based
credibilityasan
effectivetreatment
forthebroadspectrum
of alcoholics
(Monti
et al., 1989).In thecurrentstudy,however,CBT appearedto
producefewerabstinent
daysthanTSF forclientswithoutpsychopathology.
Thecomparative
advantage
ofTSF disappeared
asthelevelof clientpsychopathology
increased.
Because
there
werefew participants
with veryhighlevelsof psychopathologyin theoutpatient
study,furtherresearch
isneededontheeffectivenessof CBT and TSF in thesetypesof individuals.
Results,however,didnotsupportthehypothesized
superiority
of CBT for clientswithhigheralcoholinvolvement,
cognitive
impairment,sociopathyand supportfor drinking,nor for
womenandType B alcoholicsassuggested
in theliterature.
Althoughclientvariablesdidnotdemonstrate
strongmatchingeffects,severalclientattributes
werepredictiveof drinking
outcomes.For the aftercaresubjects,gender(male) and,to a
lesserextent, greateralcohol involvementand supportfor
drinkingwere associated
with lesssuccessful
outcomes.For
theoutpatients,
highermotivationfor changewasstronglyassociated
withbetteroutcomes,
andhigherlevelsof supportfor
drinkingwasassociated
withpoorerdrinkingoutcomes.
Characteristics
thatemerged
asimportant
for theaftercare
armmay
relatemoreto therelapseprocesssinceaftercareclientswere
generallyabstinentat the beginningof the ProjectMATCH
treatments,
whereasin theoutpatientarm predictorvariables
maybe morerelatedto theprocessof movingtowardandstabilizingabstinence
whichwastheinitialtaskof theoutpatient
treatments.
Theseeffectsforbothaftercareandoutpatient
subjectsareconsistent
with priorresearch(Skinner,1981;Timko
et al., 1993). Of clinicalrelevanceis the fact thatbothmotiva-
PROJECT
MATCH
RESEARCH
GROUP
25
tion andsocialsupportfor drinkingaremodifiablecharacteristicsandsuggest
theneedfor treatment
strategies
thattarget
them(DiClementeet al., 1992;Longabaugh
et al., 1995).
Combiningtheresultsof attribute,
treatment
maineffectand
treatmentmatchinganalyses
offersan interesting
view for alcoholismtreatmentresearch.Someclientattributesappearto
impactdrinkingoutcomes
irrespective
of typeoftreatment
and
deservefurtherresearch.With respectto treatmenteffects,
thereappeartobefew differences
in outcome
amongtheseindividuallydelivered
psychosocial
treatments
regardless
of dif-
in theefficacyof matchingtreatmentsto subjectcharacteristics and they certainlychallengethe existingview that attributeby treatmentmatchingis a key to improvedtreatment
effectiveness.
However,the matchingfindingsmustbe interpretedcautiouslysincethereareadditionalareasof matching andlevelsof complexityof matchingthatrequirefurther
investigation.
Moreover,the evidenceof matchingwith the
psychiatricseverityattributeoffersan importantand interestingareafor futureresearch.
In additionto testingfor matchingeffects,this trial of-
feringphilosophies
andstrategies,
consonant
withmanyprior
studiesin psychotherapy
research
(Beutler,1991;Hesterand
Miller, 1995;SmithandGlass,1977).In fact,it isprecisely
resultslike thesethathavebeenusedin thepastto arguefor ef-
fers the treatment field a wealth of new information
fortstoidentifya setofcommonactiveingredients
oftreatment,
or for investigating
clientattributeby treatment
matches
that
couldbringtreatmentdifferences
to light(HesterandMiller,
1995;Instituteof Medicine,1990;Miller andHester,1986).
Althoughprior researchhas indicatedthe potentialfor
matchingeffectsin alcoholismtreatment(Mattsonet al.,
1994), ProjectMATCH foundlittle evidencefor hypothesizedmatchesevenwith characteristics
thathadpreviously
produced
positivematchingresults(sociopathy,
conceptual
level,severityof alcoholinvolvement).
Thereareseveralpotentialexplanations
for thisdiscrepancy.
Oneexplanation
is
thata large-scale,
multiple-sitestudywith a largenumberof
subjects
eliminatedeffectsthatmayhavebeenidiosyncratic
or site-dependent.
Priorresearch,generally,hadfewer subjects, more drop out, less-controlledtreatments,lesscontrolledrandomization
procedures,
andweremorelikely
to be conducted
at a singletreatmentsite.Otherexplanations
include differences in the treatments or treatment modalities.
Most prior researchuseda grouptreatmentmodality,few
studiesusedthe exact treatmentsevaluatedin this trial, and
nonecomparedtheseparticulartreatmentswith oneanother
for matchingeffects.Althoughthereareotherpossibletypes
of matches,
particularlyclientby therapistinteraction
effects
(Beutler,1991),thatwerenot studiedsystematically
in Project MATCH, our inabilityto find robustattributeby treatment interactionsis consistentwith the resultsof prior
attributeby treatmentinteractionsresearchin educationand
psychotherapy(Dance and Neufeld, 1988; Smith and
for al-
coholism treatment and for psychotherapyin general.
Analysesof the therapyvideotapes,therapistcharacteristics, treatmentcomplianceand assessment
measuresused
in this trial will offer new informationto guidefuturetreatment studies.Anotherareathat requiresfurtherexploration
is the apparentbenefit gainedfrom prior inpatientor day
hospitaltreatmentby clientsrecruitedin the aftercarearm.
Althoughoutcomesappearedbetterfor aftercareclientsin
comparisonwith outpatientclients(see Figures 1 and 2),
causal inferences are difficult
because of the lack of ran-
dom assignmentto study arms. Possibleexplanationsfor
these differences are: the attrition of unmotivated
clients
before recruitmentinto the aftercarearm (since subjects
had to complete prior treatmentbefore inclusion); the
respite from alcohol exposureand consumptiongained
from a period of protectedabstinence;and the greaterintensityof treatmentreceivedby aftercareclientsjust prior
to participationin ProjectMATCH.
The ProjectMATCH ResearchGroupwill continueconductingplanneda priori and exploratoryanalyseswith this
uniquedatasetthroughanextensiveanalysisandpublication
plan.In Januaryof 1998thisdatabasewill be madeavailable
throughNIAAA for analyses
by otherqualifiedinvestigators.
Appendix: Project MATCH ResearchGroup,
CollaboratingInvestigators,CollaboratingFacilities
and Data Monitoring Board
PROJECT MATCH
RESEARCH GROUP
National Institute on Alcohol Abuse and Alcoholism
Sechrest,1991; Snow, 1991).
Prior alcoholismtreatmentmatchingstudieshave been
criticizedfor lack of methodologicalrigor (Lindstrom,
1992).ProjectMATCH carefullyaddressed
a numberof critical methodological
anddesignissuesthatoftenthreatenthe
intemalandextemalvalidityof clinicaltrials:clearlyarticulateda priorihypotheses,
successful
randomassignment,
use
of manualsfor all conditions,
monitoringtreatmentdelivery,
assessment
of treatmentfidelity, delivery of an adequate
amountof treatment,limitingattrition,andreliableoutcome
assessment.
ProjectMATCH is thelargest,statisticallymost
powerful,psychotherapy
trial ever conducted.The limited
matchingfindingsmay disappointmany who havebelieved
John P. Allen, Ph.D., Science Officer
MargaretE. Mattson,Ph.D., Staff Collaborator
Clinical
Research Units
William R. Miller, Ph.D.,P.I., andJ. ScottTonigan,Ph.D.,co-P.I.,
Universityof New Mexico,Albuquerque,
NM
GerardJ. Connors,Ph.D., P.I., andRobertG. Rychtarik,Ph.D., coP.I., ResearchInstitute on Addictions, Buffalo, NY
CarrieL. Randall,Ph.D.,P.I., andRaymondF. Anton,M.D., co-P.I.,
MedicalUniversityof SouthCarolinaandVeteransAffairsMedical Center,Charleston,SC
Ronald M. Kadden, Ph.D., P.I., and Mark Litt, Ph.D., co-P.I., Uni-
versityof ConnecticutSchoolof Medicine,Farmington,CT
26
JOURNAL
OF STUDIES
ON ALCOHOL
Ned L. Coohey,Ph.D., P.I., VeteransAffairs ConnecticutHealthcare Systemand Yale University Schoolof Medicine, New
Haven, CT
CarloC. DiClemente,Ph.D.,P.I., andJosephCarbonari,Ed.D., coP.I., Universityof Houston,Houston,TX
Allen Zweben,D.S.W., P.I., Universityof Wisconsin-Milwaukee,
Milwaukee, WI
RichardH. Longabaugh,
Ed.D., P.I., andRobertL. Stout,Ph.D., coP.I., BrownUniversity,Providence,RI
DennisDonovan,Ph.D., P.I., Universityof WashingtonandVeteransAffairsPugetSoundHealthCareSystem,Seattle,WA
CoordinatingCenter
/ JANUARY
1997
IvanhoeTreatmentCenter,Milwaukee,WI (Marion R. Romberger)
LawrenceCenter,WaukeshaMemorial Hospital,Waukesha,WI
(Fred Syrjanen,M.S., C.A.D.C.-III)
Medical University of South Carolina, Institute of Psychiatry,
Charleston,SC (JamesC. Ballenger,M.D., Director)
Metro Milwaukee Recovery Center, Milwaukee, WI (Steve
Skowlund,M.A., C.A.D.C.-III)
MilwaukeePsychiatric
Hospital,Wauwatosa,
WI (PattyPriebe,R.N.)
NorthwestGeneralHospital,Milwaukee,WI (RichardHicks)
RogerWilliams GeneralHospital,Providence,RI (Ted D. Nirenberg,Ph.D.)
Schick-ShadelHospital,Seattle,WA (JamesW. Smith,M.D., and
Michael Olsson,M.S.)
Thomas F. Babor, Ph.D., P.I., and FrancesK. Del Boca, Ph.D., co-
P.I., Universityof Connecticut,Farmington,CT
Bruce J. Rounsaville, M.D., co-P.I., and Kathleen M. Carroll,
Ph.D., co-P.I., Yale University,New Haven,CT
SinaiSamaritanHospital,Milwaukee,WI (Tom Johnston,
M.S.W.)
SoutheasternWisconsin Medical and Social Services, Milwaukee,
WI (Lawrence Neuser, President)
Southwood
CommunityHospital,Norfolk,MA (YolandaLandrau,
R.N., Ed.D., and Rhoda Stevens, R.N., C.A.C.)
Consultant
Veterans Affairs Medical Center, Charleston, SC (Bryon
PhilipW. Wirtz, Ph.D., GeorgeWashingtonUniversity,Washington, DC
COLLABORATING
INVESTIGATORS
Adinoff, M.D.)
VeteransAffairs Medical Center,Houston,TX (LaurenPate, M.D.,
and Su Bailey, Ph.D.)
Veterans Affairs Medical Center, Milwaukee, WI (Dennis Borski,
M.S.W., and Jung-KiCho, M.D.)
Su Bailey, Ph.D., Veterans Affairs Medical Center-Houston,
and Departmentof Psychiatry,Baylor College of Medicine,
DATA MONITORING
Houston, TX
KathleenBrady,Ph.D., M.D., Instituteof Psychiatry,MedicalUniversityof SouthCarolina,Charleston,SC
Ron Cisler, Ph.D., Center for Addiction and Behavioral Health Re-
search,Universityof Wisconsin,Milwaukee,WI
Reid K. Hester,Ph.D., BehaviorTherapyAssociates,
Albuquerque,
Daniel R. Kivlahan,Ph.D., VeteransAffairs Puget SoundHealth
CareSystem-Seattle,
andDepartmentof PsychiatryandBehavioral Sciences,Universityof WashingtonSchoolof Medicine,
Seattle, WA
Ted D. Nirenberg,Ph.D., Roger Williams Medical Center and
Brown University,Providence,RI
BOARD
PaulCushman,Jr., M.D., Departmentof Psychiatry,StateUniversityof New York, StonyBrook,NY
JohnFinhey,Ph.D., Centerfor Health Care Evaluation,Program
Evaluationand ResourceCenter (152), VeteransAffairs Medical Center, Menlo Park, CA
Ralph Hingson, Sc.D., Social Behavior and SciencesSection,
BostonUniversitySchoolof PublicHealth,Boston,MA
James Klett, Ph.D., Bel Air, MD
Michael Townsend, Ph.D., Division of SubstanceAbuse, Cabinet
for Human Research, Frankfort, KY
Lauren A. Pate, M.D., Veterans Affairs Medical Center-Houston,
and Departmentof Psychiatry,Baylor College of Medicine,
Houston, TX
Acknowledgments
Ellie Sturgis, Ph.D., Medical University of South Carolina,
Charleston, SC
Consultant
Larry Muenz, Ph.D., Gaithersburg,
MD
COLLABORATING
FACILITIES
CareUnit Hospital of Kirkland, Kirkland, WA (Karen PorterFrazier,R.N., andJanBigby-Hanson,M.S.W.)
CharlestonCountySubstance
AbuseCommission,Charleston,
SC
The steeringcommitteewouldlike to acknowledgetheeffortsof a working group,chairedby RobertStout,in guidingandconducting
analysesreported in this article: Frances Del Boca, Joseph Carbonari, Carlo
DiClemente,RichardLongabaugh,Scott Toniganand Philip Wirtz. We
alsothank the writing team of Carlo DiClemente,RichardLongabaugh,
GerardConnorsand RobertStoutfor their effortsin the preparation
of
draftsof the manuscript.
Notes
(BarbaraDerrick, Executive Director)
CPCGreenbriar
Hospital,Milwaukee,WI (DonaldC. Fischer,M.D.)
DePaulHospital,Milwaukee,WI (BrianE. Tugana,M.D., M.B.A.)
FenwickHall Hospital,Charleston,
SC (JohnMagill, C.E.O.)
Harris CountyPsychiatricCenter,Houston,TX (Ken Krajewski,
M.D., andTerry Rustin,M.D.)
1. Althoughthesetenvariableswereselectedasthemostpromisingfor generatingand testingmatchinghypotheses,
a numberof other variables
havebeenincludedin secondary
hypotheses
thatwill notbe thefocusof
the currentreport.Thesesecondaryhypotheses
includemeasuresof
DSM-III-R Axis I diagnostic
categories,
clientself-efficacy,alcoholdependence,
anger,deviance,socialfunctioning,antisocialpersonality
dis-
PROJECT
2.
3.
4.
5.
MATCH
order,religiosity,assertionof autonomy,anothermotivationmeasure
and severalhigherorderandglobalmatchinghypotheses,
all of which
havepotentialasmatchingvariables.
This variablewasdefinedby a clusterof interrelated
indicators
of premorbidvulnerability(e.g., a family historyof alcoholdependence)
and
currentproblemseverity(e.g., alcoholdependence
syndrome)that permittedclassification
of clientsaseitherTypeA alcoholics
(low vulnerability and moderateproblemseverity)or Type B alcoholics(high
vulnerability
andsevereproblems).
Theprocess
of formulating
a priorimatchinghypotheses
involvedinitial
literaturereviewsto identifypromisingclient attributesandtreatments
that would likely provide a basis for client-treatmentinteractions
(Longabaugh
et al., 1994).Writtenproposals
werecritiquedby the steeringcommitteeandthe mostpromisingoneswereselected
for testingin
thetrial asprimaryhypotheses.
Thesewentthroughseveraliterationsof
refiningtheirrationale,specifying
predictions,
anddeveloping
assumed
"causalchains"thatdescribedprobablemechanisms
of actionfor eachof
theproposed
client-treatment
interactions.
At eachstageof development
theseproposals
werereviewedandcritiquedby a hypothesis
reviewcommittee(R. Longabaugh
andP. Wirtz) andthesteeringcommittee,
prior
to final acceptance
by thesteeringcommittee.
TheMilwaukeeclinicalresearch
unitwasprimarilyanoutpatient
sitebut
developedanaftercarecapacityandcontributed
subjects
to boththe outpatientandaftercarearmsof the study.
Supplementary
analyses
indicatedthattherewereno significantGGTP
differences
amongtreatment
conditions
atanyof thethree(3-, 9- and15month)follow-uptimepointsfor eitherarm.
6. Two sets of covariates were examined. The first set included the baseline
valueof thecriteriondrinkingmeasureanditsinteraction
with time.The
second added the CRU effect terms as described in the text. Both sets of
covariatesyieldedsimilarresults.Only the resultsof the analysesusing
the secondset of covariatesare reportedsincethesecovariateswere
judgedtobethemostappropriate
to adjustforbothsiteandbaselinemeasurement effects.
References
ANTON,R.F. ANDBEAN,P. Two methodsof measuring
carbohydrate-deficienttransfertinin inpatientalcoholicsandhealthycontrolscompared.
Clin. Chem. 40: 364-368, 1994.
ANTON, R.F. AND MOAK, D.H. Carbohydrate-deficient
transferrinand
gamma-glutamyltransferase
asmarkersof heavyalcoholconsumption:
Genderdifferences.
AlcsmClin. Exp.Res.18: 747-754,1994.
BABOR,T.F., Reliability of the EthanolDependenceSyndromescale.Psychol. Addict. Behav. 10: 97-103, 1996.
BABOR,T.F., LONGABAUGH,R., ZWEBEN, A., FULLER, R.K., STOUT, R.L.,
ANTON, R.A. AND RANDALL, C.L. Issues in the definition and mea-
surementof drinking outcomesin alcoholismtreatmentresearch.J.
Stud.Alcohol,Supplement
No. 12,pp. 101-111,1994.
BEUTLER,
L.E. Towardspecificpsychological
therapiesfor specificconditions.J. Cons.Clin. Psychol.47: 882-897, 1979.
BEUTLER,
L.E. Haveall wonandmustall haveprizes?RevisitingLuborsky
et al.'s verdict.J. Cons.Clin. Psychol.59: 226-232, 1991.
BISHOP, Y.M.M., FEINBERG,S.E. AND HOLLAND, P.W. Discrete Multi-
variateAnalysis:Theoryand Practice,Cambridge,Mass.:MIT Press,
1975.
BOWMAN, K.M. AND JELLINEK,E.M. Alcohol addiction and its treatment.
Q.J. Stud. Alcohol 2: 98-176, 1941.
BROWN, S.D. Therapeuticprocessesin Alcoholics Anonymous.In:
McCRADY,B.S. AND MILLER, W.R. (Eds.) Researchon Alcoholics
Anonymous:Opportunitiesand Alternatives,New Brunswick,N.J.:
RutgersCenterof AlcoholStudies,1993,pp. 137-152.
BRYK,A.S. ANDRAUDENBUSH,
S.W. Applicationof hierarchicallinear
modelsto assessing
change.Psychol.Bull. 101: 147-158,1987.
RESEARCH
GROUP
27
CARBONARI,J.P., WIRTZ, P.W., MUENZ, L.R. AND STOUT, R.L. Alterna-
tive analyticalmethodsfor detectingmatchingeffectsin treatmentoutcomes.J. Stud.Alcohol,SupplementNo. 12, pp. 83-90, 1994.
CARROLL,K.M., CONNORS,G.J., COONEY,N., DICLEMENTE, C.C., DONO-
VAN, D.M., KADDEN,R.M., LONGABAUGH,
R.L., ROUNSAVILLE,B.J.
ANDZWEBEN,A. Discriminabilityandintegrityof treatments
for alcoholism:FindingsfromProjectMATCH, submitted
for publication.
CARROLL,K.M., KADDEN, R.M., DONOVAN, D.M., ZWEBEN, A. AND
ROUNSAVILLE,
B.J. Implementingtreatmentandprotectingthe validity
of the independent
variablein treatmentmatchingstudies.J. Stud.Alcohol,Supplement
No. 12,pp. 149-155,1994.
CLIFFORD,P.R. AND LONGABAUGH,R. Manual for the Administration of
the ImportantPeopleandActivitiesInstrument(availablefrom Richard
Longabaugh,Brown University,Center for Alcohol and Addiction
Studies,800 Butler Drive, Providence,RI 02906), 1991.
CONNORS,G.J., ALLEN, J., COONEY,N.L., DICLEMENTE, C.C., TONIGAN,
J.S. ANDANTON,R. Asssessment
issuesand strategiesin alcoholism
treatment
matchingresearch.
J. Stud.Alcohol,Supplement
No. 12,pp.
92-100, 1994.
COONEY,
N.L., KADDEN,R.M., LITT, M.D. ANDGETTER,H. Matchingalcoholicsto copingskillsor interactional
therapies:
Two yearfollow-up
results.J. Cons.Clin. Psychol.59: 598-601, 1991.
Cox, D.R. Regression
modelsandlife tables.J. RoyalStat.Soc.(SeriesB)
34: 187-220, 1972.
CRONKITE,R.C. ANDMOOS,R.H. Sex and marital statusin relation to the
treatmentandoutcomeof alcoholicpatients.SexRoles11:93-112, 1984.
CRUMBAUGH,
J.C. The Seekingof NoeticGoalsTest (SONG): A complementaryscaleto the Purposeof Life Test (PLT). J. Clin Psychol.33:
900-907, 1977.
CRUMBAUGH,
J.C. ANDMAHOLIK,L.T. Purposein Life Scale,Murfreesboro,Tenn.:Psychometric
Affiliates,1976.
DANCE,K.A. AND NEUFELD,R.W.J. Aptitude-treatmentinteractionresearchin theclinicalsetting:A reviewof attempts
to dispelthe"patient
uniformity"myth.Psych.Bull. 104: 192-213,1988.
DEL BOCA,F.K., ANDBROWN,J.M. Issuesin the developmentof reliable
measuresin addictionsresearch:Introductionto ProjectMATCH assessment
strategies.
Psychol.Addict.Behav.10: 67-74, 1996.
DICLEMENTE,C.C., CARBONARI,
J.P. ANDVELASQUEZ,M.M. Alcoholism
treatmentmismatching
from a processof changeperspective.
In: WATSON,R.R. (Ed.) AlcoholAbuseTreatment(Drug andAlcoholAbuseReviews,Vol. 3), Totowa,N.J.: HumanaPress,1992,pp. 115-142.
DICLEMENTE,C.C., CARROLL,K.M., CONNORS,
G.J. AND KADDEN,R.M.
Processassessment
in treatmentmatchingresearch.J. Stud. Alcohol,
Supplement
No. 12,pp. 156-162,1994.
DICLEMENTE,
C.C. ANDHUGHES,
S.O. Stagesof changeprofilesin outpatientalcoholismtreatment.J. Subst.Abuse2: 217-235, 1990.
DICLEMENTE,C.C., PROCHASKA,
J.O., FAIRHURST,S.K., VELICER,W.F.,
VELASQUEZ,
M.M. ANDROSSI,
J.S. The processof smokingcessation:
An analysisof precontemplation,
contemplation
andpreparationstages
of change.J. Cons.Clin. Psychol.59: 295-304, 1991.
DONOVAN,D.M., KADDEN,R.M., DICLEMENTE,C.C., CARROLL,K.M.,
LONGABAUGH,R., ZWEBEN, A. AND RYCHTARIK,R. Issues in the se-
lectionanddevelopment
of therapies
in alcoholism
treatmentmatching
research.
J. Stud.Alcohol,Supplement
No. 12,pp. 138-148,1994.
DONOVAN,D.M. ANDMATTSON,M.E. Alcoholismtreatmentmatchingresearch:Methodologicaland clinical issues.J. Stud.Alcohol,SupplementNo. 12,pp. 5-14, 1994.
DONOVAN,D.M., WALKER,R.D. ANDKIVLAHAN,D.R. Recoveryandreme-
diationof neuropsychological
functions:
Implications
for alcoholism
rehabilitation
process
andoutcome.
In: PARSONS,
O.A., BUTTERS,
$. AND
NATHAN,P. (Eds.) Neuropsychology
of Alcoholism:Implicationsfor
DiagnosisandTreatment,New York:GuilfordPress,1987,pp. 339-360.
EDWARDS,
O. ANDLADER,M.H. Addiction:Processes
of Change,New
York: Oxford Univ. Press,Inc., 1994.
FINNEY,J.W. ANDMOOS,R.H. Matchingpatientswith treatments:
Conceptualandmethodological
issues.
J. Stud.Alcohol.47:122-134, 1986.
28
JOURNAL
OF STUDIES
ON ALCOHOL
/ JANUARY
1997
FINNEY,J.W. ANDMoos, R.H. Theory andmethodin treatmentevaluation.
Eval. Prog.Planning12: 307-316, 1989.
FOWLER,
J. AlcoholicsAnonymousandfaith development.In: MCCRADY,
B.S. AND MILLER, W.R. (Eds.) Researchon Alcoholics Anonymous:
Opportunities
andAlternatives,
New Brunswick,N.J.:RutgersCenterof
AlcoholStudies,1993,pp. 113-135.
MCLACHLAN,
J.F. Benefitfrom grouptherapyas a functionof patienttherapistmatchon conceptuallevel. Psychother.
TheoryRes.Prac.9:
GLASER,
R.B. Matchless.'?
AlcoholicsAnonymousand the matchinghypothesis.
In: McCRADY,B.S. ANDMILLER,W.R. (Eds.)Research
on AlcoholicsAnonymous:
Opportunities
andAlternatives,
New Brunswick,
N.J.:RutgersCenterof AlcoholStudies,1993,pp. 379-395.
O'BRIEN,C.P. Improvedoutcomesfrom problem-service
"matching"
in substance
abusepatients:A controlledstudyin a fourprogram,LAP
network,submitted
for publication.
GOODMAN,A.C., HOLDER, H.D., NISHIURA, E. AND HANKIN, J.R. A discrete choice model of alcoholism treatment location. Med. Care 30:
1097-1109, 1992.
GOUGH, H.G. California PsychologicalInventory Manual, Palo Alto,
CaliL: ConsultingPsychologists
Press,1975.
HEATHER,
$., ROLLNICK,
S. ANDBELL,A. Predictivevalidityof thereadinessto changequestionnaire.
Presented
at theInternationalConference
on the Treatmentof Addiction Behaviors,SanteFe, N.M., 1993.
nESTER,R.K. ANDMILLER,W.R. (Eds.) Handbookof AlcoholismTreatmentApproaches:
EffectiveAlternatives,
2d Edition,Boston:Allyn and
Bacon, 1995.
HUNT, D.E., BUTLER,L.F., NoY, J.E. ANDROSSER,
M.E. Assessing
ConceptualLevel by theParagraph
CompletionMethod,Toronto:Ontario
Institute for Studies in Education, 1978.
INSTITUTEOFMEDICINE.Broadeningthe Baseof Treatmentfor Alcohol
Problems.
Washington,
D.C.:NationalAcademyPress,1990,pp.279-302.
KADDEN, R., CARROLL,K.M., DONOVAN, D., COONEY, $., MONTI, P.,
ABRAMS,
D., LITT, M. ANDHESTER,R. Cognitive-Behavioral
Coping
SkillsTherapyManual:A ClinicalResearch
Guidefor Therapists
Treating Individualswith AlcoholAbuseand Dependence.
NIAAA Project
MATCH Monograph,
Vol. 3, DHHS Publication
No. (ADM) 92-1895,
Washington:GovernmentPrintingOffice, 1992.
KADDEN,R.M., COONEY,
N.L., GETTER,H. ANDLITT, M.D. Matchingalcoholics
tocopingskillsor interactional
therapies:
Posttreatment
results.
J. Cons.Clin. Psychol.57: 698-704, 1989.
KEISLER,
D.J. Somemythsof psychotherapy
research
andthesearchfor a
paradigm.Psych.Bull. 65:110-136, 1966.
LAMBERT,M.J., BERGIN,A.E. The Effectivenessof Psychotherapy.
In:
BERGIN,A.E. ANDGARFIELD,
S.L. (Eds.)Handbookof Psychotherapy
andBehaviorChange(4thEdition).New York:JohnWiley & Sons,1994.
LINDSTROM,
L. Managing Alcoholism:Matching Clients to Treatment,
New York: Oxford Univ. Press,Inc., 1992.
LITT, M.D., BABOR,T.F., DEL BOCA,F.K., KADDEN,R.M. AND COONEY,
N.L. Typesof alcoholics,
II: Applicationof an empiricallyderivedtypologyto treatmentmatching.Arch.GenPsychiat.49: 609-614,1992.
LONGABAUGH,
R. The MatchingHypothesis:Theoreticaland Empirical
Status,Washington,D.C.: AmericanPsychological
Association,1986.
LONGABAUGH,R., BEATTIE, M., NOEL, $., STOUT, R. AND MALLOY, P.
The effect of social investment on treatment outcome. J. Stud. Alcohol
54: 465-478, 1993.
LONGABAUGH,R., WIRTZ, P.W., BEATTIE,M.C., NOEL, N.E. AND STOUT,
R. Matchingtreatmentfocusto patientsocialinvestmentandsupport:
18-Monthfollow-upresults.J. Cons.Clin. Psychol.63: 296-307, 1995.
LONGABAUGH,
R., WIRTZ, P.W. ANDCLIFFORD,
P.R. The ImportantPeople and ActivitiesInstrument(availablefrom RichardLongabaugh,
BrownUniversity,Centerfor AlcoholandAddictionStudies,800 Butler Drive, Providence,RI 02906), 1995.
LONGABAUGH,
R., WIRTZ, P.W., DICLEMENTE,C.C. AND LITT, M. Issues
in thedevelopment
of client-treatment
matching
hypotheses.
J. Stud.Alcohol,Supplement
No. 12,pp.46-59, 1994.
LYONS,J.P., WELTE,J.W., BROWN,J., SOKOLOW,
L. ANDHYNES,O. Vari-
ationin alcoholism
treatmentorientations:
Differentialimpactuponspecific subpopulations.
AlcsmClin. Exp.Res.6: 333-343,1982.
MACANDREW,
C. The differentiationof male alcoholicoutpatients
from
nonalcoholic
psychiatric
outpatients
by meansof theMMPI. Q. J. Stud.
Alcohol 26: 238-246, 1965.
317-323, 1972.
MCLACHLAN,
J.F. Therapystrategies,
personalityorientationandrecovery
from alcoholism.Canad.Psychiat.Assoc.J. 19: 25-30, 1974.
MCLELLAN, A.T., GRISSOM,G.R., ZANIS, D., RANDALL,M., BRILL, P. AND
MCLELLAN, A.T., KUSHNER,H., METZGER, D., PETERS,R., SMITH, I.,
GRINSON,G.P., PETTINATI, H. AND ARGERIOU,M. The fifth edition
of the AddictionSeverityIndex. J. Subst.AbuseTreat. 9: 199-213,
1992.
MCLELLAN, A.T., LUBORSKI,L., WOODY, G.E. AND O'BRIEN, C.P. An
improveddiagnosticevaluationinstrumentfor substanceabusepatients:The Addiction Severity Index. J. Nerv. Ment. Dis. 168: 26-33,
1980.
MCLELLAN, A.T., LUBORSKI, L., WOODY, G.E., O'BRIEN, C.P. AND
DRULEY,K.A. Predictingresponseto alcoholand drug abusetreatments:Role of psychiatricseverity.Arch. Gem Psychiat.40: 620-625,
1983a.
MCLELLAN, A.T., WOODY, G.E., LUBORSKI, L., O'BRIEN, C.P. AND
DRULEY,K.A. Increasedeffectivenessof substanceabusetreatment:A
prospective
studyof patient-treatment
"matching."J. Nerv. Ment. Dis.
171: 597-605, 1983b.
MARLATT, G.A., BAER,J.S., BONOVAN,D.M. AND KIVLAHAN, D.R. Ad-
dictivebehavior:Etiologyandtreatment.Ann. Rev. Psychol.39: 223252, 1988.
MATTSON,M.E. AND ALLEN, J.P. Researchon matchingalcoholicpatientsto treatments:Findings,issuesand implications.J. Addict.Dis.
11: 33-49, 1991.
MATTSON,M.E., ALLEN, J.P., LONGABAUGH,
R., NICKLESS,C.J., CONNORS,
G.J. ANDKADDEN,R.M. A chronological
reviewof empiricalstudies
matchingalcoholicclientsto treatment.J. Stud.Alcohol,Supplement
No. 12, pp. 16-29, 1994.
MILLER,W.R. Matchingindividualswith interventions.
In: HESTER,
R.K.
AND MILLER, W.R. (Eds.) Handbookof AlcoholismTreatmentApproaches:
EffectiveAlternatives,Elmsford,N.Y.: PergamonPress,Inc.
1989,pp. 261-271.
MILLER, W.R. Manual for Form 90: A structuredassessmentinterview for
drinkingandrelatedbehaviors.
NIAAA ProjectMATCH Monograph,
Vol. 5, DHHS PublicationNo. (ADM) 96-4004, Washington:GovernmentPrintingOffice, 1996.
MILLER,W.R. ANDDELBOCA,F.K. Measurement
of drinkingbehaviorusingtheForm90 familyof instruments.
J. Stud.Alcohol,Supplement
No.
12,pp. 112-118,1994.
MILLER,W.R. ANDHESTER,
R.K. Matchingproblemdrinkerswith optimal treatments.In: MILLER, W.E. AND HEATHER,$. (Eds.) Treating
Addictive Behaviors:Processesof Change,New York: Plenum Press,
1986.
MILLER,W.R. ANDMARLATT,
G.A. The Comprehensive
DrinkerProfile,
Odessa,
Fla.:Psychological
Assessment
Resources,
1984.
MILLER, W.R., TONIGAN, J.S. AND LONGABAUGH,R. The Drinker Inven-
tory of Consequences
(DrInC): An Instrumentfor Assessing
Adverse
Consequences
of AlcoholAbuse.Rockville,Md.: NIAAA, 1995.
MILLER, W.R., ZWEBEN, A., DICLEMENTE, C.C. AND RYCHTARIK,R.G.
Motivational EnhancementTherapy Manual: A Clinical Research
Guide for TherapistsTreating Individualswith Alcohol Abuse and
Dependence.NIAAA ProjectMATCH Monograph,Vol. 2, DHHS
PublicationNo. (ADM) 92-1894, Washington:GovernmentPrinting
Office, 1992.
MONTI,P.M., ABRAMS,
D.B., KADDEN,R.M. ANDCOONEY,N.L. Treating
AlcoholDependence:
A CopingSkillsTrainingGuide,New York:Guilford Press, 1989.
NOWINSKI,
J., BAKER,
S. ANDCARROLL,
K. TwelveStepFacilitationTherapy Manual: A ClinicalResearchGuide for TherapistsTreatingIndi-
PROJECT
MATCH
vidualswith AlcoholAbuseandDependence.
NIAAA ProjectMATCH
Monograph,
Vol. 1, DHHS Publication
No. (ADM) 92-1893,Washington:GovernmentPrintingOffice, 1992.
ORFORD,J., OPPENHEIMER,E. AND EDWARDSG. Abstinence or control:
The outcomefor excessivedrinkerstwo yearsafter consultation.Beh.
Res. Ther. 14: 409-418, 1976.
PISHKIN,V. ANDFREDERICK,
S. Comparison
of extentof purposeof life
amongalcoholics
andnonalcoholics.
J.Clin.Psychol.29:387-391,1973.
PROJECT
MATCH RESEARCH
GROUP.Project MATCH: Rationaleand
methodsfor a multisiteclinicaltrial matchingpatientsto alcoholism
treatment.AlcsmClin. Exp. Res.17:1130-1145, 1993.
PRoPst,L.R. The comparativeefficacyof religiousand nonreligiousimageryfor thetreatmentof mild depression
in religiousindividuals.Cog.
Ther. Res. 4: 167-178, 1980.
ROBINS,
L., HELZER,J., COTTLER,
L. ANDGOLDRING,
E. NIMH Diagnostic InterviewSchedule:VersionIII Revised(DIS-III-R), Questionby
QuestionSpecifications,
St. Louis,Mo.: WashingtonUniversity,1989.
SASINSTITUTE.SASTechnicalReportP-229,Cary,N.C.: SAS, 1992.
SHIPLET,W.C. A self-administering
scalefor measuringintellectualimpairmentanddeterioration.
J. Psychol.9: 371-377, 1940.
SKINNER,H.A. Comparisonof clientsassignedto in-patientand outpatienttreatmentfor alcoholismanddrugaddiction.Brit. J. Psychiat.
138: 312-320, 1981.
SMITH,A. SymbolDigitModalitiesTest,LosAngeles,Calif.:WesternPsychologicalServices,1973.
SMITH, B. AND SECHREST,
L. Treatmentof aptitudeX treatmentinteractions.J. Cons.Clin. Psychol.59: 233-244, 1991.
SMITH,M.L. ANDGLASS,
G.V. Meta-analysis
of psychotherapy
outcome
studies.Amer. Psychol.32: 752-760, 1977.
RESEARCH
GROUP
29
SNOW,R.E. Aptitude-treatment
interaction
asaframeworkforresearch
onindividualdifferencesin psychotherapy.
J. Cons.Clin. Psychol.59: 205216, 1991.
SOBELL,L.C. AND SOBELL,M.B. Timeline follow-back:A techniquefor
assessingself-reportedalcohol consumption.In: LITTEN, R. AND
ALLEN,J.P. (Eds.)MeasuringAlcoholConsumption:
Psychosocial
and
BiochemicalMethods,Totowa,N.J.: HumanaPress,1992,pp. 41-42.
SPITZER,R.L. AND WILLIAMS, J.B.W. Structured Clinical Interview for
DSM-III-R, PatientVersion, New York: BiGmetricResearchDepartment,New York StatePsychiatricInstitute,1985.
STOUT,R.L., WIRTZ, P.W., CARDGNARl,J.P. AND DEL BOCA,F.K. Ensur-
ing balanceddistribution
of prognostic
factorsin treatmentoutcomeresearch.J. Stud.Alcohol,Supplement
No. 12, pp. 70-75, 1994.
TIMKO, C., FINNEY, J.W., Moos, R.H., Moos, B.S. AND STEINBAUM,D.P.
The processof treatmentselectionamongpreviouslyuntreatedhelpseekingproblemdrinkers.J. Subst.Abuse5: 203-220, 1993.
TONIGAN,J.S., MILLER,W.R. ANDBROWN,J.M. The reliability of Form
90: An instrument
for assessing
alcoholtreatmentoutcome.J. Stud.Alcohol,in press.
WANBERG,K.W., HORN, J.L. AND FOSTER,F.M. A differential assessment
modelfor alcoholism:The scalesof AlcoholUse Inventory.J. Stud.Alcohol 38: 512-543, 1977.
ZWEBEN,A. ANDCISLER,R.A. Compositeoutcomemeasures
in alcohol
treatmentresearch:
Problemsandpotentialities.
Int. J. Addict.,in press.
ZWEBEN,A., DONOVAN,D.M., RANDALL,C.L., BARRETT,D., DERMEN,K.,
KABELA,E., MCREE,B., MEYERS,R., RICE,C., ROSENGREN,
D., SCHMIDT,
P., SNOW, M., THEYOS,A.K. AND VELASQUEZ,M. Issues in the devel-
opmentof subject
recruitment
strategies
andeligibilitycriteriain multisite
trialsof matching.
J. Stud.Alcohol,Supplement
No. i 2, pp.62-69, 1994.